QuasiScience News Feed

Riding the Digital Dragon

Sat, 04 Oct 2025 23:00:00 GMT

In September 2025, the Chinese State Council launched a new plan to embed artificial intelligence throughout the Chinese economy - the AI+ initiative. AI+ is a national program that combines investment, procurement, regulation and local-level execution to make AI an engine of productivity throughout the vast country.

For businesses outside China, this development is more than a geopolitical headline. It is a masterclass in how to integrate AI technologies into operations quickly, strategically, and at scale, although the lesson also comes with some warnings.

This article distills the key points from China's AI+ strategy into actionable insights for Western businesses of every size and sector.

What is China's plan for AI?

Unlike China's previous AI strategies, the AI+ strategy focuses on the importance not of research, but of application. Rather than labs, Beijing is funding demand creation - making sure industries adopt AI to raise productivity and competitiveness across the country.

The strategy focuses on:

fiscal and procurement support to speed up enterprise trials and deployments
local and municipal pilot programs to create demand and data
workforce retraining
parallel governance and labelling regimes to control risk

It's a very applied, practical approach: for China, AI is a tool to raise national productivity, not an abstract technology. This builds on China's long-stated approach in documents such as the 2017 New Generation Artificial Intelligence Development Plan.

There are clear lessons here for Governments, but why should entrepreneurs care? Because the entire business landscape is shifting:

Your suppliers, competitors, or partners in China may soon run leaner, faster operations thanks to AI diffusion
Standards and practices from AI+ funded pilots will likely influence global norms
Most importantly, the processes China are using to diffuse AI across their nation (e.g. procurement design, pilot programs, modular adoption) can be adapted for digital business transformation, even in small businesses in non-technical industries

Five Lessons for Western businesses

The Chinese AI+ approach holds important lessons not only for transforming a vast nation's economy, but also for integrating AI into any business, large or small. Here are the top five:

1. Engineer demand, don't wait for it

China isn't waiting for AI adoption to occur 'naturally' as a result of market forces. Through subsidies, municipal pilot projects, and state procurement, it is creating early demand to fuel iteration.

Lesson for business leaders:

For rapid transformation, don't passively wait for specialist staff or customers to demand AI or digital tools. Instead:

Fund internal pilots with clear KPIs
Offer incentives for business units that adopt new systems that generate proven results
Commission expert reviewers to identify where your business could genuinely benefit from AI

2. Think local and modular, not business-wide

China's central government is setting the national direction, but it is provinces and cities that are running pilots — healthcare in one city, smart grids in another.

Lesson for business leaders:

Instead of chasing a massive, enterprise-wide digital transformation (which is often expensive and unwieldy), start with modular pilots in departments or regions. A retail company, for example, could test AI-driven offline demand forecasting in one city before rolling it out nationally.

3. Adoption is not just about tech

China's AI+ aligns investment, talent training, and standards-setting, recognising that AI adoption requires more than just good technology. Great technologies can be useless if they are not properly integrated or understood, or if they are used irresponsibly.

Lesson for business leaders:

Digital tools fail when you just buy software without also preparing people and processes. Think about:

Budgeting for integration, not just licenses
Retraining staff to use new systems
Creating company-wide standards for responsible use, data governance, and safety

4. Ignore the hype

China's rhetoric around AI+ emphasises efficiency gains, for example in factory yields, the accuracy of medical diagnostics, or the optimisation of logistics. Conversely, it is not interested in abstract 'breakthroughs' or soundbite worthy 'transformations'.

Lesson for business leaders:

Whilst few industries have as much hype around them as AI does right now, businesses, like Governments, must learn to ignore it. Instead, focus ruthlessly on ROI. Hire qualified deep technical experts to identify where AI can make a real difference in your business, or start by asking yourself:

Where your teams spend a lot of time on repetitive tasks
Where you need to influence the behaviour of large numbers of customers
Where you need to make difficult decisions based on uncertain or unreliable data
Where you need to safely and reliably organise large amounts of data
Where you need to create high quality content quickly
Where your teams spend a lot of time staying on top of trends or developments
Where you need to test lots of different products/approaches

Begin where measurable efficiency gains are likely, not where hype is loudest

5. Build (the right) partnerships

China's big tech companies (Alibaba, Baidu, Huawei, etc.) are scaling AI investment in parallel with the state, creating an effective ecosystem in which AI technologies can develop rapidly.

Lesson for business leaders:

You don't need to build everything in-house. Many of the most exciting AI technologies require deep expert, multi-disciplinary teams to implement them effectively, so it is crucial to find the right partners. This includes:

Expert engineers who can not only build software or sell you SaaS products, but also integrate and customise tools so they work for your business
Traditional engineers or specialists who understand the systems you are trying to model, automate or modernise
Training providers who can upskill your teams
Regulatory and standards experts who can ensure you are not just compliant, but responsible
Deep technical consultants who can give you an honest appraisal of where AI will actually add value, not just make a nice headline

Risks to Keep in Mind

China's AI+ strategy also brings to mind three key risks that business leaders face when implementing AI technologies.

1. Beware of access challenges in regulated markets

China's plan includes governance and labelling regimes that will affect what data and systems can be used.

Lesson for business leaders:

For Western firms, localisation (data residency, local legal constructs, compliant model governance) will also be required in many markets - make sure you have processes for safe data handling, partnerships, and any joint ventures.

2. Cultural resistance is real

Here's one way in which businesses need to act very differently from states: while China is able to issue top-down mandates for AI adoption, Western companies need buy-in from employees and customers.

Lesson for business leaders:

Adoption plans must include change management and the creation of good governance to build trust.

3. Hasty adoption can be wasteful

Last but not least, it's important to note that some Chinese pilots have succeeded because of political backing, not economics. In business, we don't have this luxury.

Lesson for business leaders:

Insist on evidence-based models of likely ROI before investing in projects.

Riding the Digital Dragon

As the old Chinese adage goes: if you ignore the dragon, it will eat you; if you try to confront it, it will overpower you; if you ride it, you will take advantage of its might.

The mighty dragon that is AI+ reminds us that, while the winners in this new business era definitely won't be those who ignore AI, they also won't be leaders who just try to outcompete everyone else for the smartest tech. The key to success is making AI adoption easy, fast, and scalable.

For Western business leaders, the lessons are clear:

Engineer demand, don't wait for it
Think local and modular, not business-wide
Adoption is not just about tech
Ignore the hype
Build (the right) partnerships
Beware of access challenges in regulated markets
Cultural resistance is real
Hasty adoption can be wasteful

As part of the right strategic approach, AI tools can bring about a positive transformation in almost any business. Contact QuasiScience today if you'd like to discuss how to get the most out of these exciting new technologies.

References

China State Council release on AI+ initiative (State Council guideline, Aug 27, 2025). (State Council of China)
Stanford DigiChina translation: "A New Generation Artificial Intelligence Development Plan" (2017) — foundational planning document that frames Beijing's AI goals through 2030. (Stanford.edu)
RAND analysis: "Full Stack: China's Evolving Industrial Policy for AI" — analysis of China's use of industrial policy across the AI stack. (RAND Corporation)
Carnegie Endowment: commentary on China's intent to diffuse AI across the economy and challenges of large-scale integration. (Carnegie Endowment)
Reporting on corporate investment: Alibaba's expanded AI spending and strategic posture (August-September 2025 reporting). (Investopedia)

Monte Carlo Simulations

Tue, 21 Oct 2025 23:00:00 GMT

In today's fast-moving business environment, uncertainty is often the only certainty. Markets shift, customer preferences change overnight, and regulatory frameworks can change with every political cycle. For leaders making high-stakes decisions - whether allocating capital, setting pricing strategies, or entering new markets - traditional forecasting tools often fall short. Ordinary statistical methods using static scenarios or deterministic models fail to capture the full range of possible outcomes, and may be foxed by extreme scenarios.

One solution to this problem is Monte Carlo Simulations. (We will cover others, such as Markov Chains and Uncertainty Quantification, in later articles).

Monte Carlo Simulations

Inspired by his uncle's gambling habit, mathematician Stanislaw Ulam hit upon the idea of using repeated random sampling to model uncertainty. Monte Carlo simulations are now widely used in finance, engineering, pharmaceuticals and research, but what are they? A simple example of a Monte Carlo simulation is a simulation to calculate the area of an geometrical shape for which you don't know the formula. Using the Monte Carlo method, you would inscribe the shape in a rectangle (for which it's easy to calculate the area), then scatter a random distribution of points over the rectangle. The proportion of points that land inside the shape give you a good idea of the proportion of the area of the rectangle that is taken up by the shape.

Case study: infrastructure Portfolio Optimisation

Our client was an innovative energy company investing in a €400 million portfolio of renewable projects, which required them to manage multiple types of uncertainty including:

Intermittent production: Solar output fluctuates with cloud cover, time of day and seasonal changes. Wind power varies hourly and seasonally.
Shifting market conditions: Prices for electricity are influenced by demand spikes, fuel costs, and regulatory changes.
Portfolio effects: Interactions between assets—such as wind farms in different regions—can amplify or dampen risks.

Our client's team had decades of experience in the energy sector, but their existing tools lacked the flexibility to handle the complexity that comes with renewable energy.

Monte Carlo simulations were a crucial part of the solution. For example, when modelling energy production from wind, the rectangle from our previous example is analogous to possible wind speeds, and the random points are analogous to samples of wind speeds at different times of year/day. The area of the abstract shape is analogous to an estimate of expected wind speed over the year (which then gives expected power generation in uncertain conditions).

Bringing together several Monte Carlo simulations, as well as unique models from our 200+ proprietary code libraries, we helped our client understand:

Production variability for several classes of renewable assets.
Volatility of market prices.
Aggregated results across portfolios, reflecting how diversification can mitigate or magnify risks.

Then integrated these into a system which provided automated bid price optimisation, so that our client could confidently make offers balancing competitiveness and risk.

Why are Monte Carlo simulations important?

Monte Carlo simulations are valuable in any situation where you need to translate uncertainty into actionable insight. They allow leaders to:

Optimise decision-making: test strategies under different conditions and select options that balance risk and reward.
Quantify risk with confidence: Instead of vague risk labels, Monte Carlo simulations provide a data-driven probability of success or failure under different conditions.
Plan for extremes: By exploring tail scenarios — rare but impactful outcomes — Monte Carlo simulations allow businesses to build resilience and avoid catastrophic surprises.

The result is a more disciplined, evidence-based approach to strategic decision-making, which is particularly critical in environments where stakes are high and volatility is the norm.

Here are a few more examples:

Finance: Assess portfolio risk, stress-test capital plans, and price complex contracts.
Supply chain: Plan inventory under uncertain demand, evaluate supplier risk, and optimize logistics strategies.
Marketing and Sales: Forecast revenue under variable adoption rates, test pricing strategies, and optimize promotional spend.
Mergers and acquisitions: Quantify deal risk, simulate synergies, and stress-test assumptions.

Governance and adoption

As ever, it's not all about the tech. To maximise the value of Monte Carlo simulations, business leaders should consider these questions:

What are your decision-critical uncertainties? Only you know which few variables drive the most risk or opportunity. Modelling every detail rarely adds value.
Is this tool usable? Insist that engineers provide tools that integrate with existing workflows, and present insights in formats you can use.
How does this tool complement human judgement? Monte Carlo simulations provide probabilities, not guarantees. Leaders need to use outputs to inform strategy.
How should we iterate this tool? In rapidly-evolving industries, tools should be regularly updated to reflect new priorities. Ensure your tools have through-life support.
How do I build confidence in this tool? Even the best tool is useless if it is not trusted. Complement new tech with the right training and governance.

Smart decisions in uncertain times

Uncertainty is inevitable, but chaos is optional. Tools such as Monte Carlo simulations enable leaders to see not just what might happen, but how likely it is, what drives it, and how to respond strategically in an unpredictable world.

But it's not all about the tech. The best engineering teams will ensure simulations are integrated with your existing systems, scale with your company, and are widely adopted, thanks to effective training and governance.

Get in touch today so we can help you model your toughest decisions and turn uncertainty into an opportunity.

Digital Twins

Thu, 09 Oct 2025 23:00:00 GMT

Digital twins - mathematical models of physical systems - are one of the most powerful emerging technologies. By enabling organisations to simulate complex products or processes before committing the resources required to make real-world prototypes, digital twins can cut costs, accelerate innovation, and reduce environmental impacts. They have applications in almost every field, from engineering and energy to pharmaceuticals and healthcare. However, successful deployment requires not only strong technical teams, but also leaders who take the right strategic approach.

What is a Digital Twin?

Digital twins can be very simple. For example, a single computer chip linked to a sensor in a pipe, running an equation to set off an alert when a water tank is about to overflow, is a basic digital twin. It uses data and a mathematical model to simulate what is going on in the tank and visualise this for users.

Although many digital twins are far more complex than this, they all have the same four elements:

Data: collected from sensors (e.g. the one in the pipe) or pre-existing datasets (e.g. the dimensions of the tank).
Model: equations or algorithms that express the relationship between inputs and results (e.g. the equation linking the volume of the tank and the rate of water inflow to when it will overflow).
Simulation: a computational platform continuously using sensor data to make a prediction using the model (e.g. the software that runs the equation and sets off an alert when the tank is nearly full).
Visualisation: Dashboards and interfaces that allow engineers, managers, or decision-makers to understand the results of the twin (e.g. the screen showing the alert)

Competitive Advantage in the Age of AI

This simple concept gives rise to exciting use cases across almost every major sector of the economy. Leaders who have mastered the art of using digital twins are already gaining a significant competitive advantage.

To give just a few examples:

For healthcare, Pfizer are using digital twins to predict how compounds will interact with biological systems, reducing reliance on animal testing. Meanwhile, Philips are developing patient-specific digital twins of hearts that let cardiologists simulate treatment outcomes and personalise care plans.
In the energy sector, Siemens are using digital twins for predictive maintenance of power plants, reducing downtime by 10%; they are also using this technology to optimise placement of wind turbines and increase yield.
Turning to aerospace, NASA are simulating deep space missions using digital twins, to anticipate mission failures and ensure that expensive equipment is not wasted.
When it comes to the automotive sector, BMW are running digital twins of automated production lines, significantly reducing planning time; while Volvo are developing vehicle twins to monitor fleet performance and facilitate remote updates.
Digital twins are big in logistics: DHL are running warehouse twins to optimise the routes of robots and the layout of shelves, improving picking efficiency; while Maersk are building digital twins of global shipping routes to improve their resilience in the face of disruption.
In retail, Walmart are using digital twins of shops and refrigeration units to reduce emergency maintenance by 30%.
And digital twins are even improving our cities: Singapore's Virtual Singapore project is a full-scale city twin for testing traffic flow, energy demand, and disaster response; while Helsinki have also created a city twin to model noise, pollution, and building energy consumption.

Five secrets engineers wish leaders knew

Despite their enormous potential, poorly implemented digital twins can be expensive and ineffective. Even small digital twin pilots require significant investments in talent, data and architecture, so it is crucial to ensure projects are not derailed by the challenges that digital twin engineers face at every stage of the development process. So how can business leaders reap the benefits of digital twins, without risking expensive project failures?

1. Ask the right questions

Returns on investment in digital twin infrastructure are significantly more likely if leaders start with the right questions. Leaders keen to explore digital twins but unsure where to start should ask themselves:

Where in my organisation are we spending a lot of time or money on prototyping or testing products or processes?
Where are we under pressure to reduce waste or emissions?
Where is the length of our innovation cycle holding us back?
Where are we managing risks that arise from complex systems e.g markets, supply chains, or regulatory environments?
Where are our operations subject to a high failure rate e.g equipment failure or missed deliveries?

Answers to these questions will give a good indication of where digital twins could generate value for your business, as a foundation for a discussion with expert engineers.

2. Demand effective integration

Unfortunately, many digital twin projects fall down not because of engineering flaws in the twins themselves, but because of a failure to integrate with the company's wider systems. A twin that cannot exchange data with Enterprise Resource Planning, supply chain platforms, Internet of Things sensors, or design software, risks becoming just an expensive visualisation tool. Leaders should insist on interoperability from the outset, ensuring that twins can both consume and generate data across the enterprise. This means aligning digital twin initiatives with existing data strategies, cloud infrastructure, and integration standards, so the insights generated can actually drive decisions and actions.

3. Invest in security

In their enthusiasm to implement exciting new technologies, technical teams sometimes forget that digital twins often bring together sensitive operational, financial, and even personal data, making them an attractive target for cyberattacks.

As always, breaches are a major business risk. Therefore, it's essential that leaders treat digital twin environments as critical infrastructure, embedding cybersecurity controls such as identity management, encryption, and network segmentation from the outset, and committing to regular testing and monitoring in alignment with GDPR.

4. Ensure stakeholder buy-in

Surprisingly, the most common reason digital twins fail is not technical. Instead, digital twin projects most often fall down due to a lack of trust from the users and the organisation more widely. Experienced employees have good reasons to distrust digital twins, worrying that they will degrade standards or create avoidable errors. For this reason, clear governance is crucial. From the outset, leaders should consider:

How clearly models express reasoning and uncertainty
Accountability for input data quality
Transparency about how digital twins work (they must never be black boxes)
A collaborative model-design process, bringing together both traditional engineers/experts and digital specialists
Through-life processes for assessing and validating twins.

5. Choose the right partners

Digital twins sit at the intersection of engineering, data, and operations, so multi-disciplinary teams are central to their success. Traditional engineers or specialists who understand the real-world systems being modelled, and data scientists who can model and analyse the systems, must work together closely. While the former group will often come from within your organisation, it can be challenging to choose the latter.

As a guideline, leaders should look for digital twin engineers who can solve problems at every stage of the digital twin engineering process. It is worth asking how they will address these issues:

Data must be effectively integrated from all the different sources involved; cleaned effectively when it may be unreliable or low quality; and kept secure. Systems must also be able to handle any latency (delay) if data is being collected in real time; and large volumes of data when necessary
Models should strike a balance between being complex enough to be useful but simple enough to work with limited processing power; draw on an advanced, inter-disciplinary understanding of the real-world system; evolve with the real-world system; and capture uncertainty
Simulations need to have sufficient computing power; synchronise effectively with live systems; and, where necessary, integrate multiple simulations (if the twin is for an entire system, rather than just one component)
Information should be displayed clearly and intuitively for non-technical users, and integrate with existing systems in the organisation

At QuasiScience, Our Innovate UK and AWS award-winning team of engineers builds specialised, evidence-based digital twins that offer real competitive advantage. Drawing on our 200+ proprietary code libraries and our relationships with top Universities, we research brand new, bespoke solutions for every project, which offer up to 80% better returns than industry-standard alternatives.

Tech Savvy Leaders Test Decisions Scientifically

Digital twins are no longer a futuristic concept — they are a practical tool already reshaping industries, and giving leaders a powerful way to de-risk decisions by exploring possibilities and testing solutions before committing significant resources.

But, as ever, the technology itself is only part of the story. Leaders must ask the right questions, ensure integration with other systems, invest in adequate security, build trust through effective governance, and partner with the right mix of technical and domain experts. Organisations that approach digital twins as strategic capabilities — rather than isolated experiments — will gain the most from their deployment.

In short, digital twins offer a rare opportunity to cut costs, accelerate innovation, and reduce environmental impact all at the same time. Leaders who seize this opportunity with discipline, vision, and the right partners will not only future-proof their operations, but also gain a lasting competitive advantage.

Affordable Advanced Simulations

Mon, 13 Oct 2025 23:00:00 GMT

For decades, advanced simulations have been the preserve of aerospace giants, F1 teams, and government laboratories equipped with multimillion-dollar supercomputers. Meanwhile, small and mid-sized businesses - the true engine of innovation in most economies - were locked out. They either faced prohibitive costs for private computers, endured long queues for access to national facilities, or, worst of all, simply gave up and relied on guesswork.

But, thanks to new research by QuasiScience, in partnership with Cranfield University and Masters student Premkumar Bet, this situation has now changed. Our team of expert engineers tested whether accurate, repeatable, and cost-effective simulations like the ones we used in F1 could run on scalable cloud systems, and opened the door to more inclusive, flexible, and sustainable innovation.

Why Computational Fluid Dynamics matters

Among the many types of simulation, Computational Fluid Dynamics (CFD) - which predicts how liquids and gases behave under different conditions - plays an outsized role in shaping modern industries. We certainly used it a lot in F1, but aerospace engineers, automotive designers, and even healthcare innovators also rely on CFD to understand how airflow, temperature, turbulence, and aerodynamic forces will change as they change their designs. It's crucial for designing quieter aircraft, more fuel-efficient vehicles, better ventilated buildings, and safer medical devices.

CFD is significantly cheaper and faster than traditional physical testing, but it's computationally very demanding. On top of this, the most valuable insights often emerge when CFD is combined with thermal analysis, structural mechanics, and control systems; and heavily automated to avoid manual data-entry errors. That creates an enormous strain on traditional computing systems.

To address these demands, many organisations turn to high performance computing (HPC) clusters - essentially, networks of powerful machines that split a problem into parts and solve them in parallel. This approach can shrink a simulation that would take weeks on a single desktop down to one that takes only hours.

But there are problems with clusters too:

Concentration of capacity: North America holds a disproportionate share of global HPC infrastructure. Access for companies elsewhere is limited.
Capital intensity: Setting up a private cluster often requires millions in upfront investment, as well as ongoing spending on specialised teams to maintain it. This simply isn't workable for smaller companies.
Environmental footprint: HPC clusters consume massive amounts of energy, and generate significant carbon emissions.

The result? Advanced simulations remained the preserve of the few, leaving countless innovators underserved.

Our research approach

At QuasiScience, we wanted to change all this. We decided to test whether accurate, repeatable, and cost-effective simulations could be run on affordable, scalable cloud systems. So we partnered with Cranfield University, a world leader in aerospace research, and Masters student Premkumar Bet, to explore this challenge.

We decided to design and deploy a scalable system to support parallel CFD simulations, that ran not on a supercomputer, but on the cloud (specifically, the widely-used Amazon Web Services). We also set up software to automatically assign different part of the jobs to different cloud computers so they work together smoothly; built a step by step system to run the entire simulation automatically, from setting up the model to running the calculation to collecting the results; and tested our results against existing systems.

For benchmarking purposes, we ran several classic CFD cases that have been widely studied in the literature, and we compared our results to results from Cranfield University's Crescent HPC cluster.

The outcomes were as we expected in terms of accuracy, but exceeded our expectations in terms of cost. Without any particular optimisation, the costs we measured were three times lower than Cranfield's traditional system for running the same exact simulation. And, because the cloud system can scale up or down seamlessly, we could do as many jobs in parallel as we wanted, lowering turnaround times for all the cluster users.

Why it matters

The implications go far beyond engineering departments. Affordable, cloud-based simulations will create new opportunities across industries. This development means:

Democratization of innovation: Start-ups and SMEs can now run simulations that were once reserved for multinationals or well-funded research institutions.
Agility and speed: Cloud clusters can scale up or down instantly. Companies can accelerate R&D timelines, respond faster to market shifts, and seize first-mover advantage in competitive sectors.
Sustainability gains: As big cloud providers move toward renewable energy, by using cloud-based clusters, businesses can align simulation practices with ESG goals and regulatory requirements.
Cost efficiency: Using cloud-based computing, businesses will pay only for the computing power they need - whether it's an occasional spike in demand or continuous development cycles. Alternatively, hybrid models could combine in-house infrastructure with cloud capacity for optimal efficiency.
Geographic accessibility: Cloud-based clusters can be made available globally, bypassing regional shortages of supercomputing infrastructure.

On top of this cloud-based simulations have arrived at a critical moment.

R&D costs are rising: companies face mounting pressure to innovate while controlling spend.
ESG imperatives are growing: Investors and regulators are demanding proof of lower carbon footprints.
Competition is global: Emerging-market players are seeking ways to compete with established Western and Asian incumbents.

Forward-looking executives should see cloud-based HPC as a truly strategic capability.

Which industries should care?

The potential for cloud-based high performance clusters extends well beyond our initial aerospace example. Some of the industries likely to benefit most include:

Automotive: Simulating aerodynamic drag, battery cooling, and crash scenarios for electric vehicles.
Pharmaceuticals: Running molecular simulations to accelerate drug discovery, particularly for underfunded diseases where traditional supercomputing is unaffordable.
Energy: Optimising wind farm layouts or simulating heat transfer in next-generation nuclear systems.
Built environment: Designing buildings with improved airflow, natural ventilation, and reduced HVAC energy use.
Consumer products: Rapidly testing and optimising designs for sports equipment, wearables, or appliances, which are not high value enough to merit traditionally computational fluid dynamics simulations.

How to implement cloud-based HPCs

There are three key steps to capture the benefits of high-performance clusters:

Identify pilots: Think about simulation/testing use cases that did not previously represent value for money, but may do so at a three times lower cost.
Choose the right partners: Find technical partners who can not only run a cloud-based high performance cluster, but also integrate it effectively with your systems and advise you on how to use it strategically.
Institutionalise the shift: Train teams and revise governance to embed cloud-based simulations into your R&D cycle.

The organisations that invest early will gain a twofold advantage: lower costs and faster insights today, plus the ability to scale innovation rapidly tomorrow.

Conclusions

Cloud computing levels an important playing field. Simulations which once required multimillion-dollar infrastructure can now be done inexpensively, without compromising on accuracy, speed, or sustainability.

Cloud-based HPC will enable faster innovation in critical fields (like drug development for underfunded conditions), reduced costs for start-ups and SMEs with brilliant ideas but limited resources, and greener operations for every business that makes the transition.

Keen to take advantage of this groundbreaking research? Contact us today.

London Defence Tech Hack 2025

Sat, 24 May 2025 23:00:00 GMT

[London, UK] - [March 25, 2025] - QuasiScience, a growing Simulation and Data Science company, attended the London Defence Tech Hack 2025, an event dedicated to exploring innovative solutions for the defence sector.

Event Details

Dates: May 17-18, 2025
Location: Royal Military Academy Sandhurst, UK

Program Overview

During the intense two-day event, over 250 of the best young engineers, start-ups and industry leaders descended on the Royal Military Academy Sandhurst for London Defence Tech Hackathon 2025 to solve real life defence problems from the UK UK Ministry of Defence. The hackathon will provide a platform for developers, designers, and industry experts to work together and learn from each other.

Outcomes

Statement from Marco Ghilardi, Managing Director of QuasiScience:

It was a deeply enriching experience to be part of the hackers team at the Royal Military Academy Sandhurst, an important centre for military training with great historical significance. The event gave us several insights into how technology can enhance operational capabilities. A big thank you to the personnel at Sandhurst for their hospitality and support, and to the sponsors and organisers for making this event possible.

About QuasiScience

QuasiScience mission is help businesses become more successful through the use of advanced Numerical Simulations, Automation, and Data Science. With a team of talented industry experts, researchers, and developers, QuasiScience is pushing the boundaries of process optimisation through the use of Mathematics.

Related Resources

Official Linkedin Page

Contacts

PR Team

Partnership with London Lounge Radio

Sun, 20 Apr 2025 23:00:00 GMT

[London, UK] - [March 31, 2025] - QuasiScience, a growing Simulation and Data Science company, is partnering with London Lounge Radio ("LLR"), an organisation specialising in the curation of exceptional musical events and art exhibitions.

This partnership will help both companies accelerate their grow through strategic resource and knowledge sharing. LLR core expertise lies in the creation of engagement and design of unique experiences; through our partnership they will be able to speed up the development of certain features of their platform. Meanwhile, QuasiScience will benefit from LLR expertise in audience engagement, allowing us to better connect with our users and clients.

Statement from Alexander Maffei, Director of London Lounge Radio:

Thanks to the team at QuasiScience we already feel more confident in creating more ambitious events and we will soon kick off the development of a new platform that will allow us to better connect with our audience. We are excited to work with QuasiScience and look forward to the innovative solutions we can create together.

Statement from Marco Ghilardi, Managing Director of QuasiScience:

We are excited to partner with London Lounge Radio, an organisation that shares our passion for innovation and creativity. This partnership will allow us to apply our expertise in data science to help LLR create even more engaging and memorable experiences.

About QuasiScience

About London Lounge Radio

London Lounge Radio is dedicated to fostering spaces where up-and-coming musicians and artists can connect with audiences looking for authentic, innovative experiences.

Contacts

PR Team

Hackathon for the Creative Industries

Sat, 11 May 2024 23:00:00 GMT

[London, UK] - [May 12, 2024] - QuasiScience, a growing Simulation and Data Science company, is partnering with HOST at MediaCityUK to support a hackathon focused on developing innovative AI solutions for the creative industries.

This exciting event will bring together talented developers, designers, and creative thinkers from across the region to collaborate and develop novel AI applications for various creative fields. The hackathon will provide participants with access to QuasiScience expertise in developing custom machine learning implementations and deep industry knowledge from researchers at the University of Salford and MediaCityUK.

QuasiScience's CEO, Marco Ghilardi, expressed enthusiasm for the event, stating:

We are thrilled to partner with the HOST for this initiative. The focus on craftsmanship and control over the creative process has made the creative industries more resistant to the adoption of new technologies at scale. This hackathon will provide a platform for artists and innovators to co-develop solutions and test the boundaries of what's possible. We are excited to see the creativity and ingenuity that will emerge from this event.

Hackathon Details

Dates: May 15 and 16, 2024
Location: MediaCityUK, Salford, M50 2NT
Theme: AI for the Creative Industries

About QuasiScience

QuasiScience mission is help businesses become more successful through the use of advanced Numerical Simulations and Data Science. With a team of talented industry experts, researchers, and developers, QuasiScience is pushing the boundaries of process optimisation through the use of Mathematics.

About HOST

HOST is Salford's Home Of Skills & Technology. It is an innovation hub at the heart of MediaCityUK and offers unique environment to learn, grow and succeed. HOST aims to bring companies and start ups under one roof to foster innovation and help individuals learn the skills to take on more technical roles.

About the University of Salford

The University of Salford is a renowned academic institution with the mission of broadening access to education. The University's Immersive Technology department is a leader in the field of AI research applied to media and production.

Contacts

PR Team

AI-driven Film Editing Software

Sun, 14 Jul 2024 23:00:00 GMT

[Bristol, UK] - [July 15, 2024] - QuasiScience, a growing Simulation and Data Science company, will be co-developing Virtuoso AI with Nulight Studios.

Virtuoso AI is a suite of powerful video editing tools that bridge the gap between generative AI and the film and television industry. Our MVP will automate the identification and replacement of unwanted objects in video: a time-consuming task for Visual Effects (VFX) Artists that often requires meticulous frame-by-frame work. By automating this process, filmmakers can produce high-quality content more efficiently, streamline post-production, and significantly reduce costs.

This tool will be particularly beneficial for natural history documentary makers in the South West, removing unwanted objects like lens dirt, car or town lights, and radio collars on animals to ensure pristine footage.

Key advantages of the AI-powered object-replacement tool include seamless integration with existing professional film editing and VFX software, support for industry workflows and file formats, and enhanced IP security.

This innovative tool will showcase the potential of the Virtuoso AI platform, opening new avenues for creativity and efficiency in film and television production and establishing Nulight as pioneers in AI-driven video editing.

Unlike other AI platforms, integrity and transparency is at the core of Virtuoso AI. The models are trained using legally sourced and licensed films, with a list of sources made publicly available. Additionally, the platform ensures customer footage cannot be collected or used for training.

About QuasiScience

About Nulight Studios

Nulight Studios is a leading UK provider of motion picture film scanning, restoration and digital remastering services to the broadcast and film distribution market.

Related Resources

MyWorld Press Release
Nulight Press Release

Contacts

PR Team

Fashion Innovation in Florence

Fri, 15 Dec 2023 00:00:00 GMT

[Florence, IT] - [December 15, 2023] - QuasiScience, a growing Simulation and Data Science company, will be attending the first Demo day of the Italian Fashion-tech accelerator StyleIt.

Event Details

Dates: December 15, 2023
Location: Manifattura Tabacchi, Florence, ITA

About QuasiScience

About StyleIt

StyleIt is the FashionTech accelerator of the CDP National Accelerator Network: an initiative of CDP Venture Capital SGR together with Startupbootcamp and GELLIFY.

Contacts

PR Team

Smart marketing

Mon, 20 Sep 2021 00:00:00 GMT

The missing 50%

As retailers compete in an environment where customer expectations for personalisation are higher than ever, even small improvements in the relevance of recommendations can translate into large revenue increases. Marketing is therefore typically one of the largest investments for retailers, often around 7-8% of annual revenue. But despite this substantial investments, many of them still rely on historical performance, simple attribution models, or only instinct to allocate budgets. Retailers face increasing pressure to maximise returns on their marketing spend. As 19th-century retailer, John Wanamaker famously put it:

Half the money I spend on advertising is wasted; the trouble is I don't know which half.

-- John Wanamaker

Our client, an S&P500 luxury retailer, was using simple recommendation rules (such as 'most bought items') that failed to incorporate data on user behavior. This meant they could only offer static, low-context suggestions that neither reflected individual preferences nor adapted to the shopper's intent. They client knew that millions in potential revenue were being wasted, but lacked the modelling capability to capture this opportunity.

What's wrong with traditional marketing models?

Most marketing mix models are good at explaining what has happened historically, and optimising for the retailers' key objective, whether that's increased revenue or reduced markdowns. However, they rarely account for wider economic conditions such as inflation, consumer confidence, interest rates or seasonality, and typically optimise against only a single objective. In reality, market conditions are changing more quickly all the time, and marketing leaders must balance multiple competing priorities simultaneously: increasing sales, protecting margins, maintaining brand value, acquiring new customers and growing market share.

A research-based approach

For our client, drawing on our 200+ proprietary code libraries, QuasiScience developed a bespoke optimisation platform that combines state-of-the-art statistical modelling with up-to-date macroeconomic data.

The key capabilities include:

Macroeconomic forecasting: covering inflation, consumer confidence, seasonal effects and wider economic indicators.
Integration of diverse data sources: including historical data and commodity price trends.
Multi-objective optimisation: Simultaneous optimisation for revenue, profitability, return on investment and other strategic objectives, rather than a single metric.
Scenario planning: Modelling of thousands of different situational and budget-allocation scenarios to allow marketing teams to compare options before committing spend.
Actionable recommendations: on how marketing budgets should be distributed across channels and campaigns to maximise commercial performance.

The solution was also integrated seamlessly into the client's existing data infrastructure, enabling rapid adoption without disrupting established processes.

Benefits for both customers and the business

The project delivered significant commercial value:

A 3% uplift in sales across various categories, totalling an additional £2m over the first year of the pilot.
An 80% cost reduction compared to the former off-the-shelf solution
Faster, evidence-based decision-making, as merchandising, digital, and marketing teams now have clear evidence on which types of recommendations perform best.

Applications across industries

However you use them, one thing is clear: cutting-edge recommendation engines are crucial for competitive advantage in an increasingly data-driven retail landscape.

Who should use marketing mix modelling?

We'd recommend investing in marketing mix optimisation as soon as customer acquisition emerges as a cost of >£10k per year. This is not just a crucial capability for retailers: it is also relevant for financial services, telecommunications, travel, media and e-commerce businesses.

By using advanced simulation and optimisation techniques, organisations can move beyond reporting marketing performance to actively engineering better outcomes.

QuasiScience offers marketing mix optimisation as either a fixed-price off-the-shelf tool or a bespoke service. Get in touch today to discuss your requirements.

Beyond Big Budgets

Fri, 06 Jun 2025 00:00:00 GMT

As visual effects and animation production increasingly rely on high-fidelity rendering, long processing times can cause significant delays. Working with Nulight studios, QuasiScience developed a novel cloud-based system to accelerate rendering tasks, which gives smaller studios access to high-performance rendering without significant infrastructure investments.

Hidden Costs of Visual Effects

Rendering high-quality visual effects for film and animation requires a lot of computing power. For studios without extensive hardware, this creates bottlenecks that delay production, and costs can be prohibitive.

Our client, Nulight studios, was developing a VFX plugin, which meant they needed to render large sequences to test and compare various options for the tool efficiently. But, using traditional rendering technology, this was unworkably slow and costly.

Alternative Approach

Drawing on our collaborative research with top UK universities, QuasiScience decided to accelerate the testing process by moving Nulight's rendering to the cloud. This solution meant:

Accessibility: Pay-per-use cloud-based solutions can enable high-performance rendering capabilities without significant upfront investment in infrastructure.
Scalability: Cloud-based resources are used only when they are needed, and can be scaled up and down rapidly.
Computing power: Just like a traditional, hardware-based HPC, cloud-based rendering can perform tasks in parallel, enabling high-throughput processing.
Rapid model validation: Nulight Studios was able to complete their first AI-based VFX plugin in only 3 months - a process that could otherwise have taken twice as long.

This system provided the client with not only affordable power, but also speed and flexibility, making it possible for a relatively small studio to produce a truly innovative VFX product.

Open Access Beta

We are currently working on a ready-to-use beta model of our cloud rendering infrastructure to provide studios of all sizes with scalable, high-throughput rendering capabilities.

Reach out to our team via our contact page if you'd like to be part of the Beta test, or to learn more about how we can help accelerate your visual effects production.

Automated sales prep

Sat, 31 Jan 2026 00:00:00 GMT

In fast-paced B2B environments, sales success depends on a detailed understanding of customers. But many sales teams spend more time searching for information than speaking to their clients! Even with the best CRM in the world, it is time consuming for teams to refresh their memories of previous conversations and get the information they need to make a sale.

Our client, a global industrial manufacturer, asked us to solve this problem. We developed a chat-based agent that allowed sales teams to access relevant context instantly, leading to a 65% reduction in prep time, with no tedious manual data entry into a CRM tool.

By saving so much time on data-entry, this tool meant we could focus on the most crucial part of CRM - understanding human motivations and developing strategies.

— Client Sales Director

The data trap

Sales organizations are under increasing pressure to deliver personalized, data-backed campaigns, but most existing CRM systems are simply not up to the task. In reality, reps avoid using these tools due to the time required and the complexity of using the system, leading to data gaps, and missed opportunities.

Our client asked us for a solution that would allow sales reps to work in an intuitive way, using simple easy-to-use chat environments, but still gain the depth and precision that comes from the most detailed CRMs.

Conversational CRM

Drawing on cutting-edge automation and ML research, QuasiScience built a conversational agent that integrates directly with the company's IT architecture. It offers:

Conversational interface: Sales reps simply message the tool to receive insights, summaries, or data drawn from relevant project documents.
Automated context retrieval: Our tool identifies the client or topic being discussed and automatically finds relevant documentation.
Automatic data recording: Data is automatically captured and organized, ensuring accuracy without manual input.

Impact

Our tool had an immediate on efficiency and sales performance:

65% reduction in prep time
Access to consistent, complete data, captured automatically from conversations
Sales teams focused on strategy, not data-entry and retrieval
A 5% increase in sales in the first six months

Our tool turned sales preparation from a chore into a painless process, so sales teams could focus on relationships, not record-keeping.

Conversational intelligence is not just for sales teams. The same framework can be used for automatic retrieval of testing documentation; maintenance records; or order histories; meaning there are applications in every business area from R&D to customer service.

Does your team have a data retrieval challenge? Get in touch to find out how we can help.

Battery capacity

Mon, 20 Sep 2021 00:00:00 GMT

Adding battery storage to renewable energy plants can significantly increase the value of renewable generation, by making power available when it's most needed, and priced most highly by grid operators. However, determining whether the investment is worthwhile is far from straightforward.

The financial return from a battery depends not only on its cost and capacity, but on how it is operated. Market prices, grid services and other market mechanisms determine the optimal control strategy, while the way the battery is operated directly affects its expected lifetime.

Ortus Climate Mitigation, a renewables company, was considering adding battery capacity across its solar portfolio, with estimated capital expenditure of approximately £15M. But they needed a way to evaluate the return on different configurations before committing the investment.

Modelling the Asset and the Market Together

A conventional financial model can estimate the return from a battery, but only once its operating behaviour is known. Conversely, a technical battery model can determine an optimal control policy without necessarily revealing whether the resulting investment makes financial sense. QuasiScience connected the two. We built an engine for control-policy design and connected its outputs directly to a Discounted Cash Flow (DCF) framework, allowing engineers to evaluate technical configurations and immediately understand their financial implications. The tool enabled Ortus to: Optimise control policies, determining how batteries should operate under different market conditions. Model asset lifetime, accounting for the relationship between operating behaviour and expected battery degradation. Test different configurations, allowing engineers to explore alternative battery capacities and plant designs. Connect engineering and finance, translating technical performance directly into financial returns. Compare investment opportunities, providing a consistent framework for evaluating where capital would generate the greatest return.

£15M Put to the Test The analysis revealed that the technically attractive investment was not necessarily the financially attractive one. Modelled the economics of battery capacity across Ortus's solar portfolio. Connected operational control strategies directly to financial returns. Demonstrated that, under UK market conditions, there was almost no financial advantage to making the proposed battery investment. Identified Morocco as a significantly more attractive destination for the company's capital. Ortus subsequently pursued the Moroccan investment instead.

What is your next £15M decision really worth? Complex investment decisions often sit at the intersection of engineering, markets and finance. Looking at only one of those dimensions can produce the wrong answer. QuasiScience combines simulation, optimisation and financial modelling to help organisations understand the real economics of complex technical investments. Talk to us before committing capital to your next big engineering decision.

Freedom from the inbox

Mon, 20 Sep 2021 00:00:00 GMT

How QuasiScience’s bespoke automations save thousands of hours of staff time and enhance operational accuracy

In corporate environments, almost all teams face the same productivity drain: email. Valuable data is buried in unstructured messages, slowing response times and increasing the risk of human error.

Our client, an S&P500 listed retailer, asked QuasiScience to address this challenge better than they could with off-the-shelf tools. We built a suite of automated inbox bots to extract and verify data, then re-route or file e-mails appropriately, saving hundreds of hours a year and making staff’s work more rewarding.

QuasiScience automated the boring, repetitive parts of my job and enabled me to focus on more interesting and valuable work

Pilot user

E-mails, e-mails, e-mails

In global retail operations, teams handle thousands of e-mails daily. There are messages from suppliers, from logistics partners, from stores and regional offices, many of which contain time-sensitive information such as delivery updates, pricing data, or compliance documentation.

Manually extracting the data in these emails, validating content, and forwarding to the right teams consumes thousands of staff hours every year. And often, it also introduces avoidable errors, or leads to information being missed.

Our client wanted to automate routine email actions and improve productivity, without disrupting existing IT systems or compromising security. So they called in the engineers!

From boredom to business impact

Drawing on our 200+ proprietary code libraries and our extensive research into new automation techniques, QuasiScience developed a custom inbox automation system which could act on incoming emails in real time. The tool:

Extracts data automatically: Pulls out data and attachments and store them instantly in the appropriate databases, eliminating manual re-entry.
Routes emails intelligently: Categorises and redirects messages to the correct departments, leading to faster internal response times.
Enhances security: Identifies anomalies and potential phishing before messages reach teams.
Validates information: Finds data that seems incorrect and asks for confirmation.

Unlike off-the-shelf tools, it is also smoothly integrated with the client’s existing infrastructure, with the option of working in either Google Workspace and Microsoft 365.

Improved efficiency and wellbeing

The tool had immediate operational and financial benefits:

70% reduction in email handling time
More than 3000 hours per year saved on average for each staff member
Reduced operational and security risk thanks to automated validation of data
Faster decision cycles across sales and operations
Employees report improved job satisfaction

Why choose bespoke automations?

Unlike more widely-available tools, bespoke tools can:

Handle operational complexity: they don’t just sort messages: they extract, validate, flag anomalies, and update downstream systems in real time.
Work across your entire organisation, eliminating manual re-entry, cross-checks, and error-prone forwarding at every stage of your business processes.
Work natively in whatever environments are relevant to you
Include phishing, anomaly detection and other first-line defences
Meet enterprise-grade data-handling standards

We’d be happy to discuss your bespoke inbox automation needs.

Saving coastlines

Mon, 20 May 2024 00:00:00 GMT

Across the world, coastal erosion threatens infrastructure, ecosystems, and cultural heritage sites. As part of our social impact research programme, to help address this problem, QuasiScience built an intelligent system capable of detecting, analysing, and predicting erosion patterns by combining satellite imagery with historical weather and ocean data to highlight high-risk areas, identify where erosion was likely to progress fastest, and prioritise locations requiring intervention.

Coastal zones provide ecological and socioeconomic services but sea-level-rise will worsen coastal erosion and the cost of inaction is substantial.

— World Bank, 2023

The high cost of erosion

The costs of coastal erosion are vast - not just financially, but also environmentally and culturally. Annual losses to housing, transport infrastructure, biodiversity, and heritage sites are escalating across Europe. Yet most monitoring remains reactive: decision-makers frequently intervene only after erosion is visibly advanced or damage has already occurred.

As part of QuasiScience's programme of environmentally-friendly research, we decided to build an early-warning intelligence platform capable of predicting erosion risk before critical thresholds are reached.

Next generation solutions

We built a technical prototype focused on modelling shoreline change precisely and scalably.

It featured:

Satellite Imagery Analysis: Multi-temporal shoreline extraction to detect long-term regression patterns and sediment movement.
Historical Weather & Climate Integration: Including data on wind, wave energy, storm frequency, tidal cycles, precipitation and seasonal variability.
CFD-Based Coastal Dynamics Simulation: Modelling how water movement, storm surges and local topography accelerate or slow erosion.
Predictive Erosion Risk Modelling: Combining imagery, climate data and hydrodynamic simulations to forecast where problems are most likely to emerge.

The framework can support governments, insurers and ecological agencies with a forward-looking tool to predict high-risk erosion zones.

As climate-related risk rises, tools like this that merge satellite intelligence, climate history and physics-based modelling will become essential for national resilience planning.

Do you have an environmental challenge you'd like to address with new technologies? Contact us to see if we can help.

Customer research

Mon, 20 Sep 2021 00:00:00 GMT

Professional networking sites contain an enormous amount of useful information for sales teams, but extracting and organising it can be slow and repetitive. Finding the right people, recording promising contacts and understanding professional backgrounds often requires sales researchers to move constantly between profiles and manually filling in or updating spreadsheets on internal systems.

A global telecommunications company's sales team asked QuasiScience to make this process more efficient.

Turning profiles into usable data

Drawing on our 200+ proprietary code libraries, QuasiScience engineered a custom browser extension that integrated directly into target customers' professional profiles.

The extension allowed our client's sales team to:

Bookmark profiles, creating a structured collection of relevant users.
Generate summaries, rapidly distilling the most relevant information from profiles.
Extract connections, turning information from professional networks into structured data for further use.
Use a reusable research workflow, eliminating the need to manually copy information between LinkedIn and other systems.

80% time savings

The extension reduced the time the sales team spent on online research by >75%. This freed them up to make more calls, and generate more leads for our client.

Integrating approaches

The best automation doesn't always require replacing an existing system with 'AI'. Often it is better to intelligently combine machine learning and deterministic systems, to make existing tools dramatically more powerful.

Do you have a repetitive data-gathering process you'd like to automate? Talk to us today.

Digital Lung

Mon, 20 Sep 2021 00:00:00 GMT

Inhaled medicines save lives every day. But they are notoriously difficult to develop, because calculating the correct dosage requires an understanding of where and how particles will deposit inside the lungs. The behaviour of an inhaled drugs depends on complex interactions between the formulation, airflow and anatomy, making physical testing expensive and time-consuming.

Anemos Therapeutics, a medical technology company developing new approaches to inhaled drug design and administration, needed a reliable simulation platform that would allow its experts to investigate drug deposition inside the lungs and advance potential drug candidates more quickly.

From Formula 1 to a digital lung

There is a surprising amount of overlap between our Formula 1 heritage and the technology needed to understand what happens to inhaled drugs inside a lung: both require modelling the movement of air and particles through a highly complex biological system. The simulations need to be sufficiently detailed to provide useful insight, while also being rapid and robust enough to support real-world decisions.

QuasiScience worked with Anemos to understand the computational fluid dynamics (CFD) simulations required to model drug deposition within the lungs, to conduct validation, and to prepare the technical documentation needed to demonstrate that the approach was scientifically credible.

Our platform provides:

CFD-based lung simulations, modelling airflow and particle behaviour within the respiratory systems of different types of people.
Drug deposition modelling, allowing experts to investigate where inhaled material accumulates.
Rapid iteration, allowing simulations to support ongoing drug-design decisions.

A solid research foundation

QuasiScience's platform gave Anemos the technical evidence and validation framework needed to take its technology forward, helping them secure a €500k investment and win an innovation competition sponsored by AstraZeneca. We are continuing to work with them as they continue their mission to make better, cheaper inhaled drugs available to all who need them.

A new frontier for drug development

The new frontiers in drug development that have become possible due to 21st century technologies come with growing business risk. Pharmaceutical companies' limited resources must be allocated across competing assets, indications and development programmes, often with significant uncertainty around clinical and commercial outcomes. Manufacturing and supply networks span hundreds of interconnected processes, making it challenging to optimise cost, capacity and resilience while maintaining quality and reliable patient supply. And companies face growing pressure to reduce emissions, energy consumption and waste while maintaining product quality, compliance and continuity of supply.

Thanks to advanced simulations like this one, they can test promising new drug candidates, delivery systems, and approaches to manufacturing safely and affordably, before making high-risk investments in real world testing. This brings better medications and patient outcomes ever closer.

Efficient competitor tracking

Mon, 20 Sep 2021 00:00:00 GMT

In retail, pricing, product positioning and competitor activity change constantly. To stay competitive, brands need accurate, up-to-date intelligence from rival websites, particularly on pricing and resales. However, collecting this information at scale is notoriously difficult.

Our client, a consortium of luxury retailers, asked QuasiScience to develop a faster, more economical alternative to existing market intelligence web scrapers.

The Challenge of competitive intelligence

Traditional web scrapers often require bespoke development for every website, creating significant maintenance costs whenever sites change their structure. This is a particular challenge when it comes to resale sites, on which not every entry is relevant. More recent ML-powered tools promise greater flexibility, but normally process every page using large language models, making them extremely compute intensive and costly.

Our clients asked for a scalable, flexible, cost-effective platform.

Smarter Architecture

Drawing on our 200+ proprietary code libraries and extensive data science research, QuasiScience engineered a hybrid web intelligence platform that combines deterministic models with machine learning tools, and automatically updates to reflect changes to the websites it is targeting. Key capabilities include:

Cross-site data collection: monitoring multiple competitor websites through a single platform.
Hybrid extraction architecture: using deterministic scraping wherever possible and AI only where additional reasoning is required.
Automatic adaptation: to changing website layouts.
Structured data pipelines: transforming unstructured web content into consistent datasets.
Scalable infrastructure: capable of monitoring large product catalogues with minimal operational overhead.

Commercial Impact

The project had an immediate impact on both costs and business processes:

Reduced annual software costs from approximately £100,000 to £20,000 for each retailer.
Consolidated multiple workflows into a single platform.
Increased flexibility as new websites of relevance to be added quickly.
Delivered reliable, structured market intelligence.

Applications Across Industries

Intelligent web scraping platforms have applications wherever organisations depend on external information. Retailers can monitor competitors; financial firms can collect market intelligence; and procurement teams can compare supplier pricing. Bespoke platforms that combine deterministic engineering with ML approaches can make data collection faster, more scalable and significantly more cost-effective.

If your organisation depends on timely, high-quality external data, get in touch today.

Efficient visual search

Mon, 20 Sep 2021 00:00:00 GMT

Large scale image recognition is increasingly important across a variety of industries: from retailers who want to enable customers to find products more easily; to manufacturers who need to identify components and defects; to insurers assessing claims; and healthcare providers working with medical imaging.

Our client, Facture, aims to make top-of-the-range art history research accessible to everyone, with their beautiful online archive. But they needed a faster, more accurate way to identify and retrieve images from their collection, which contains millions of works. Existing tools were simply too slow, and too expensive to deliver Facture's vision of democratising art research.

To build a better solution, they picked QuasiScience, winner of the 2024 prize for Amplifying Imagination: AI in the Creative Industries, sponsored by the BBC, AWS, Innovate UK and the Digital Catapult.

Why is image recognition still so difficult?

Searching for an image is a fundamentally different engineering challenge from searching for text. An image contains no keywords or labels, so a computer must first determine what visual information is important before it can compare one image with another.

Traditional systems rely on manually entered metadata, such as titles, tags or descriptions. However, metadata is often incomplete, inconsistent or subjective. Two people may describe the same object in completely different ways, while visually similar objects may share identical descriptions despite being distinct.

Modern Machine-Learning based approaches solve this by converting every image into a high-dimensional numerical representation. Images with similar visual characteristics occupy nearby positions in this vector space, allowing similarity searches that are far more flexible than keyword matching. These flexible models can recognise objects despite changes in scale, lighting, orientation or partial occlusion.

However, this flexibility comes at a cost. Generating and comparing the numerical representations requires significant computational resources, and applying deep learning inference to every search increases latency, infrastructure costs and energy consumption.

Our client needed a solution capable of searching more than 12 million images while maintaining the responsiveness users expect and avoiding the cost of running large AI models for every request.

The hybrid solution

Based on our original research, QuasiScience built a hybrid image recognition engine that combines deterministic approaches with machine learning. Rather than treating every search equally, our novel platform selects the most appropriate technique for each query. It offers:

Feature-based image matching for high-quality images: For clear, high-quality images, deterministic algorithms detect distinctive local features and geometric key points. These can be matched extremely quickly, with minimal computational overhead.
Semantic similarity search: When images are degraded, cropped, poorly lit or visually ambiguous, deep learning models enable the platform to identify visually similar objects even when pixel-level comparisons would fail.
Intelligent orchestration: A decision layer automatically determines which approach to use.
Scalable search infrastructure: The efficient architecture enables collections of multiple millions of images to be searched at the time.

Art is not what you see, but what you make others see

The project transformed Facture’s operations:

They were able to replace a commercial image recognition subscription costing £25,000 per year for the capability to search 100,000 images with a lower-cost one-off investment that enabled them to search more than 12 million.
Search times were significantly reduced by avoiding unnecessary use of compute.
Retrieval accuracy was improved for poor-quality, incomplete and visually challenging images.
The platform was scalable, and capable of supporting future growth without escalating licensing costs.

QuasiScience are more like part of the team than an external supplier. They are super smooth – very fast and thoughtful.

-- Christian Huhnt, CEO, Facture

What do you need to search?

Hybrid image recognition has applications wherever organisations need to search or classify large visual datasets. The relevance of this technology is only going to expand: as consumers demand to be able to search for products with a photograph; as autonomous systems need to interpret their surroundings; and as professionals doing fieldwork across sectors need to identify relevant items visually.

Get in touch today if you need to deliver large-scale high-fidelity image search without sacrificing speed or investing in an expensive subscription.

F1, faster

Mon, 20 Sep 2021 00:00:00 GMT

In our old field - Formula One - engineering teams operate under extreme time pressure. Design and manufacturing processes must be fast enough to support rapid iteration, and even small errors can create costly delays.

When some old friends from F1 called us back to help them beat the pressure and improve their processes, we jumped at the chance.

Turning Processes into Systems

Across industries, many engineering processes contain dull, repetitive tasks that nonetheless require significant specialist input, or depend on complex chains of data moving between design, manufacturing and operational systems. However, with the technologies available today, even complex processes can often be automated.

Working with our F1 friends and drawing on our own bespoke research, QuasiScience developed and tested new approaches to automate processes and improve the speed and reliability of design and manufacturing.

Projects included:

Design automation, reducing manual work in complex engineering processes.
Pressure line routing, automating a labour-intensive engineering task.
Data management, improving how engineering information moved through workflows.
3D printing optimisation, identifying opportunities to reduce manufacturing errors.

Six Months of Engineering Time Recovered

The pilots demonstrated that targeted automation could deliver significant gains even in highly specialised engineering environments. We reduced 3D printer errors by 20%, and saved approximately 6 months of engineering time.

What could your engineers achieve with better processes?

There's nothing to be embarrassed about - even F1 engineers waste time on repetitive processes, when they should be solving difficult problems. But with the latest automation technologies, this can change.

Financial security

Mon, 20 Sep 2021 00:00:00 GMT

For financial services businesses, technology infrastructure can be make or break. It must be secure, reliable and flexible enough to support employees wherever they work, while controlling the cost and complexity of the underlying technology estate.

An established London asset management firm asked QuasiScience to conduct a complete redesign of its internal infrastructure and network, creating a more secure and sustainable foundation for their work.

New Foundations

The existing infrastructure presented both operational and security challenges. Our award-winning engineers focused on:

Technology infrastructure - Redesigning the firm's internal technology systems to make them more reliable, scalable and easier to manage.
Network - Restructuring the company's internal network to create a more secure and dependable foundation.
IT infrastructure - Modernising the technology used to run the firm's most important systems and workloads.
Systems management - Introducing modern tools, including Kubernetes and Flatcar, to make it easier to run and manage technology consistently.
Monitoring - Using Grafana to give the team a clearer view of how their systems are performing and quickly identify problems.
Remote working - Improving the technology and security needed for employees to work effectively from outside the office.

Lower Costs, Lower Risk

The redesigned infrastructure delivered both financial and operational benefits. Our efficient architecture enabled a 70% reduction in software licensing costs, whilst also allowing employees to work remotely and improving security.

Could your technology estate be doing more for less?

Infrastructure decisions can lock organisations into unnecessary costs and security risks for years. Redesigning these foundations to make them more secure, flexible and cost-effective is often an even better investment than bringing in the latest 'AI' tools.

Talk to us today about your entire tech strategy. Let's build competitive advantage from the ground up.

Simulated testing

Mon, 20 Sep 2021 00:00:00 GMT

In today's hyper-competitive economy, even market leading construction companies are under pressure to reduce the cost, time and waste involved in testing new components.

Our client, FIP MEC Srl, a global civil engineering company specialising in high-performance components for large-scale infrastructure, was developing a new generation of bridge dampers — critical components that prevent dangerous oscillations and structural fatigue in long-span bridges. To meet safety standards, traditionally, designing and certifying these dampers requires multiple rounds of physical prototyping and destructive testing, each of which consumes significant time, resources, and materials. They were looking for a smarter, data-driven way to conduct testing.

Digital twins: not your online alter ego

To solve this problem, drawing on our decades of engineering experience, QuasiScience developed a digital twin — a mathematical model of the physical system used to simulate complex products or processes before committing the resources required to make real-world prototypes. We:

Aggregated and analysed thousands of data points from the client's historical design and testing archives. This included information about geometry, material properties, mass distribution, and vibration responses of all the previous designs.
Calculated a comprehensive mathematical model that captured how different structural parameters influence performance outcomes.
Created a digital twin which could accurately simulate whether a proposed new damper design would meet certification standards.
Visualised test results for the client's engineers, before a single prototype was built.

This allowed engineers to explore multiple variations virtually, assess their probability of passing safety tests, and refine their designs for a fraction of the cost and time required previously.

Digital twins offer competitive advantage

The digital twin transformed FIPMEC's damper design process. Now that their engineers can validate concepts digitally, they are saving €100,000 per test, per damper, as well as reducing delivery times and waste. And most importantly, the QuasiScience solution provides greater confidence in safety and performance outcomes, helping meet stringent certification requirements faster and more sustainably.

Applications across industries

Digital twins are no longer a futuristic concept - they are a practical tool already reshaping industries, and giving leaders a powerful way to de-risk decisions by exploring possibilities and testing solutions before committing significant resources.

But, as ever, the technology itself is only part of the story. Organisations that approach digital twins as strategic capabilities — rather than isolated experiments — will gain the most from their deployment. Read our white paper on "digital twins: five secrets engineers wish leaders knew" to learn more.

Faster Design and Prototyping

Mon, 26 Sep 2022 23:00:00 GMT

FIP MEC Srl is an Italian company with a long tradition in developing system that are safety critical. These components are expensive to manufacture and always need to pass stringent tests that measure the. Each time a prototype fails at testing the company needs to bear the cost of material, labour, and disposal. We engaged with them in an investigation to demonstrate how a digital twin built on previous data would be able to drastically reduce development time and costs.

Challenges

FIP MEC Srl had a state of the art laboratory but lacked the infrastructure needed to clean up and organise the unstructured results produced during testing and the formal certification process. This caused a bottleneck in the development process, as engineers had to spend a lot of time manually analysing data and running simulations to understand the results of tests and to design new prototypes. Furthermore, in cases when the test data was lost or corrupted, the company had no way to recover it, leading to costly delays and the need to repeat tests.

Solution Design

Given the challenges faced by FIP MEC Srl, we designed a solution that would allow them to clean up and organise their data, and to build a digital twin that could be used to simulate the behaviour of their components under different conditions. The digital twin was built using machine learning algorithms that were trained on the historical data produced during testing. This allowed us to create a model that could predict the behaviour of the components under different conditions, and to identify potential issues at design time, before the components were manufactured and tested.

Implementation

Data Collection

The first step in the implementation process was to collect and clean up the data produced during testing. We worked closely with the engineers at FIP MEC Srl to understand the data and to identify the relevant features that could be used to train the machine learning algorithms.

The raw data was composed of a series of reports coming from different testing machines in the internal lab and from the external certification process. We developed a pipeline that was able to extract the relevant information from these reports and to organise it in a structured format that could be used for keeping the Engineering team informed and for training machine learning models.

Digital Twin

Once the data was collected and organised, we trained a machine learning model to create a digital twin of the components. Given that this model was going to be used in a safety critical context, we focused on building a model that was not only accurate but also interpretable, so that engineers could understand the predictions and the underlying reasons behind them. We built the model by combining different sub-models that were able to capture different aspects of the behaviour of the components, such as their mechanical properties, their response to different loads, and their failure modes. This allowed us to create a digital twin that was able to simulate the behaviour of the components under different conditions without losing interpretability.

Deployment

FIP MEC had an IT Team and on premise cluster to run most of the software for internal use. Hence, we designed the system in such a way that could be delivered both on prem and cloud.

Summary

This was a landmark project for QuasiScience: it represented our first engagement outside the UK and the first consulting project. As a byproduct of this work, our team developed core capabilities to take a product from inception to market in a very short time frame and with a clear focus on the client.

The software originally developed to answer FIP MEC's business needs has seen many updates and iterations over time. We called it is now available to all businesses that look for a tailored and effective experiment tracking software and results and accelerating the development process.

The Future of KYC

Mon, 20 Sep 2021 00:00:00 GMT

Know Your Customer (KYC) processes are becoming increasingly complex. Financial institutions must comply with evolving anti-money laundering (AML) regulations while processing growing volumes of customer data from a wide range of internal and external sources. For many organisations, this results in labour-intensive workflows, duplicated effort, and lengthy onboarding times.

Our client, a consortium of Italian banks, wanted to modernise their KYC process by reducing manual administration and improving the consistency of compliance decisions. Existing workflows required analysts to gather information from multiple commercial databases, compare conflicting records, document discrepancies, and compile evidence into regulatory dossiers. These repetitive tasks consumed valuable analyst time and increased the risk of delays and human error. Meanwhile, off-the-shelf tools on the market automated only part of the operational lifecycle, and failed to integrate effectively with internal data.

Engineering compliance

Enter, QuasiScience KYC.

We are currently beta-testing an intelligent KYC platform. Unlike existing tools, it:

Automates the entire operational lifecycle - QuasiScience KYC automatically pre-fills customer information, cross-references multiple authoritative data sources, identifies any inconsistencies requiring human review, assigns follow-up actions, tracks decisions and regulatory deadlines, and compiles comprehensive audit-ready dossiers.
Integrates directly with corporate data sources, ensuring customer records remain up to date as new information becomes available.
Integrates macroeconomic and financial intelligence alongside traditional compliance data, offering richer context for risk assessment, particularly when clients face changing conditions.
Offers security against emerging, AI-enabled threat vectors

Beyond finance

We are currently testing this platform in the Italian financial system, but plan to develop versions for other regulated industries, including legal and accounting firms, real estate agencies, crypto-asset service providers, and all businesses subject to customer due diligence requirements.

By automating repetitive compliance activities while preserving human oversight for complex decisions, QuasiScience KYC enables financial institutions to reduce operational costs, accelerate customer onboarding, improve auditability, and strengthen regulatory compliance.

Would you like to participate in the Beta test? Contact us today.

Futuristic Film Editing

Sat, 31 Jan 2026 00:00:00 GMT

Machine learning has opened up extraordinary opportunities for the visual effects industry. Labour-intensive tasks like object removal and scene cleanup can now be automated. Working with Nulight studios, a leading UK provider of motion picture film scanning, restoration and digital remastering services to the broadcast and film distribution market, QuasiScience developed an award-winning tool for object removal and imperfection correction, which won the 2024 award for Amplifying Imagination: AI in the Creative Industries, sponsored by the BBC, AWS, Innovate UK and the Digital Catapult.

A game-changer for us at Nulight Studios

— CEO, Nulight Studios

Common Challenges in Film Editing

Video-editing is a creative profession, but traditional post-production work requires extensive manual frame-by-frame editing to remove unwanted objects or imperfections. It's a particular problem for natural history documentary makers, for whom removing unwanted objects like lens dirt, car or town lights, and radio collars on animals can take up the majority of editors' time, reducing their availability for more creative tasks.

By automating this process, filmmakers can produce high-quality content more efficiently, streamline post-production, and significantly reduce costs.

AI At Work

Our team of expert engineers developed a novel VFX plugin which enabled:

AI-powered object removal: automatic detection and removal of unwanted objects and imperfections in video frames.
10x quicker editing: no need to fill empty spots after object removal
Seamless integration: a plug for existing software pipelines, which didn't disrupt other work.
Improved visual quality: consistent, high-quality results across diverse types of footage.
Scalability: a repeatable, automated process that can be applied to multiple projects.
Transparency: trained using legally sourced and licensed films, with a list of sources made publicly available.

The plugin combines high-performance processing with intuitive controls, and frees up creatives to focus on creativity.

This work won a £50,000 award for Amplifying Imagination: AI in the Creative Industries, sponsored by the BBC, AWS, Innovate UK and the Digital Catapult.

The Future of Film Editing

This tool is a milestone not only in the video editing industry, but also in the tense conversation about creativity and AI. At QuasiScience, we believe that AI does not exist to replace creative jobs or lessen them, but to free them up from dull, repetitive tasks and enable them to focus on the more inspiring, imaginative work that they actually want to do.

Do you have a dream VFX plugin or image/video-processing software you'd like to develop? If so, reach out to us at QuasiScience - we'd love to help you bring it to life.

If drugs could learn

Mon, 20 Sep 2021 00:00:00 GMT

A particularly promising idea in 21st century pharmaceuticals is that of precision medicine - medicine that draws on continuous feedback to change dosages or drug types to improve patient outcomes. However this is complicated because it depends on understanding how biological systems change over time, including multiple systems that operate on very different timescales, from rapid changes (like an allergic reaction) to much slower underlying dynamics (e.g. aging).

Our client, a household name Pharmaceuticals company, wanted to invest in precision medicine at scale. But internal projects had stalled over challenges in data management and model design.

Modelling Across Multiple Timescales

Modelling a complex system like a human body requires difficult trade-offs - modelling everything quickly become computationally expensive and intractable; but simplifying the model too aggressively can make it inaccurate.

QuasiScience worked with our client's internal experts to develop an approach which could:

Represent multiple timescales, capturing both rapid and slower system dynamics within the same modelling framework.
Approximate complex interactions between systems, reducing the need to explicitly calculate every time step.
Forecast sensor measurements, allowing models to predict what sensors would observe under different scenarios.
Maintain accuracy, balancing computational efficiency with meaningful predictions

Faster Models, Better Forecasts

Based on our work, the client have now commenced an entire research programme on adaptive drugs that respond to changes in the patient's body. New modelling technologies are creating bold frontiers in pharmaceuticals and patient outcomes.

Intelligent information management

Tue, 01 Sep 2026 00:00:00 GMT

As infrastructure portfolios expand, so does the volume of project documentation. Every planning decision, engineering drawing, environmental assessment, contract and regulatory submission must remain accessible for decades.

Our client, Ortus a leading renewable energy developer and asset manager, asked QuasiScience to future-proof its knowledge management by engineering an intelligent, auditable platform capable of organising, retrieving and contextualising millions of documents spanning the entire lifecycle of its assets. They wanted more than better filing: they wanted a genuinely automated system that would enable engineers and compliance teams to focus on higher-value tasks.

Information is an asset and a risk

Renewable energy assets typically operate for 30-40 years or more. Throughout that time they generate an ever-growing body of technical reports, planning applications, environmental surveys, contracts, financial records, maintenance logs and regulatory correspondence.

Yet despite the strategic value of this information, it is often fragmented across multiple systems, folders and file formats. Documents are named inconsistently, duplicated repeatedly and become increasingly difficult to locate as portfolios grow. Engineers spend valuable time searching for information instead of acting on it, while compliance teams face unnecessary risk when critical evidence cannot be found quickly.

Engineering organisational memory

In close collaboration with the people who would be using the tool, QuasiScience engineered an intelligent platform that understood the relationships between projects, assets, people and documents.

Our solution consolidated decades of unstructured information into a single searchable environment, to deliver fast, reliable retrieval while maintaining complete traceability. It offers:

Natural language search, allowing users to ask questions in plain English instead of relying on filenames or folder locations.
Intelligent document classification, automatically organising technical, financial and regulatory information according to project context.
Context-aware retrieval, surfacing the most relevant documents rather than simply matching keywords.
Complete auditability, enabling every document, decision and supporting evidence to be traced throughout the lifecycle of each asset.
Scalable architecture, designed to support continually expanding infrastructure portfolios without degrading performance.

Better Decisions, Faster

The new platform delivered measurable operational improvements:

90% average reduction in time spent searching for documents.
Near-instant retrieval of engineering, regulatory and commercial information.
Faster, evidence-based decision-making across project development, operations and compliance.
Greater transparency for investors, regulators and stakeholders through complete audit trails.
Reduced operational risk by ensuring critical knowledge remained accessible throughout the lifecycle of every project.

"You guys rock." — Remy Marino, CFO, Ortus

How much could you save?

Whether you're managing infrastructure assets, engineering projects, legal documentation or corporate records, your teams should be able to find the answers they need in seconds, not hours.

If your organisation has complex information management requirements, get in touch today, and discover how intelligent knowledge management can transform your organisation.

21st century CAD

Sat, 15 Jun 2024 00:00:00 GMT

The things we engineers love most - modern cars, planes and other highly engineered objects - have one drawback. They require a complex, collaborative design process involving engineering teams from all over the world. Current Computer Aided Design (CAD) tools have struggled to adapt to this because of the traditional point & click interface, a large, complex CAD Kernel and reliance on powerful workstations.

Working with Kioko, a groundbreaking Franco-British startup, QuasiScience has revolutionised this process by developing an integrated CAD and testing system for engineers. The new platform alleviates many of the current limitations of CAD to answer the needs of today's engineers. The platform is the equivalent of Github and Visual Studio for 3D designs, enabling not only object visualisation, but also comprehensive testing.

This moves CAD into the 21st century

-- London Tech Week, 2024

Evolving needs

All engineers are familiar with the traditional design cycle: defining user requirements, developing ideas, planning and creating a solution, and rigorous testing. But this step-by-step description is no longer effective. The highly collaborative, international, and complex process of modern design requires rapid, iterative feedback, and increasingly uses digital simulation rather than physical testing.

Our client, Kioko, was developing a next generation CAD platform, and wanted it to reflect these changes. They decided to enhance their tool by enabling users to evaluate the performance of components at the same time as designing them, using the latest CFD mathematics.

Pushing the boundaries

Based on our own experiences of the day-to-day frustrations of an engineer, QuasiScience developed a high-performance prototype, featuring:

High-speed infrastructure: Built entirely in Rust, providing fast, memory-safe execution for simulation pipelines.
Smooth integration: Connecting design points exported from the CAD tool directly to CFD software to evaluate component performance.
Immediate feedback: using machine learning models to accelerate feedback, trading minimal accuracy when total precision was not required.

An ongoing relationship

Kioko and QuasiScience are continuing their partnership to continue developing and refining this innovative platform to meet the needs of engineers across industries. The goal is to make it the go-to tool for engineers, enabling them to design and test in a single, seamless environment. This will not only speed up the design process but also enable engineers to design and test better, more innovative products.

Optimal Store Layouts

Mon, 20 Sep 2021 00:00:00 GMT

For luxury retailers, a physical store is a major cost, but not necessarily a major generator of revenue. Customer flows tend to be smaller than in budget stores, and many who browse will not buy. However, the positioning of products, the flow and experience of customers as they move through the space and the amount of time shoppers spend in different areas can all influence whether purchases are made.

Our client, an S&P 500 luxury fashion retailer, wanted to understand whether its existing layouts were making the best possible use of the available space. Rather than relying on instinctive judgement, they asked QuasiScience to take a scientific approach.

How Customers Really Shop

Although store managers can observe customer behaviour, it would be impossible for them to capture all the interactions between thousands of customers, products and areas of a store in a scientific way. However, almost all luxury stores already have CCTV which provides a continuous record of what happens on the shop floor. The challenge is only to turn that footage into useful operational intelligence.

Our team of expert engineers conducted bespoke research on the latest computer vision techniques, then developed a system which connected to the retailer's existing CCTV. After appropriately obscuring customer identities to protect their privacy, we analysed customer behaviour and movement throughout the store and simulated how these would change with alternative layouts.

The system enabled the retailer to:

Understand how shoppers actually navigate the store.
Identify high and low-traffic areas, revealing which parts of the floor were attracting customers and which were being overlooked.
Model customer behaviour, using observed patterns to simulate alternative layouts.
Optimise product placement, testing how changing the position of products and fixtures could affect customer flow and sales.
Evaluate layouts before investing in changes, rather than relying solely on physical trial and error.
Maximise sales, selecting layouts based on scientifically predicted commercial performance rather than simply aesthetic preference.

Data is money

Our scientific analysis showed that relatively small changes to the physical environment could alter how customers moved through the store and interacted with products. After simulating and selecting the optimal layout, sales in the store where this was tested improved by 6%.

Are you making the most of your space?

You probably have CCTV already - it generates enormous amounts of behavioural data, which can be processed securely, but most retailers aren't making the most of it.

Our world-class engineers can combine computer vision, simulation and optimisation to turn that data into better decisions, so you can understand what's happening in your stores, test what could happen, and engineer environments that deliver better profits and customer experience.

How much more could your stores sell if every square metre was optimised?

Scientific pricing

Sat, 31 Jan 2026 00:00:00 GMT

As global markets becomes subject to ever-faster shifting customer demand, the pressure is on for retailers to make smarter, faster pricing decisions. Our client, an S&P 500 fashion brand, was struggling to keep prices aligned with changing market conditions.

QuasiScience created a practical automated pricing model that helped them adjust prices confidently, incorporating currency fluctuations, stock levels, and expected demand. This increased revenue by 4% compared to off-the-shelf tools.

FX and regional demand are exerting a significant pressure on margins

— Jean‑Jacques Guiony, CFO of LVMH, an industry leader

Fashion Fades

The fashion industry, and in particular fashion's eCommerce landscape, is becoming harder to navigate. Prices are updated rapidly and global currency movements affect the real value of every transaction. For retailers selling across borders, small shifts in exchange rates or demand can have a massive impact on profits.

For a company as big as our client, this was a huge challenge: prices set one day no longer made sense the next, and differing stock levels across regions meant some products sold out too quickly while others were overproduced.

But Maths is Eternal

Drawing on our 200+ proprietary code libraries of advanced simulation and optimisation models, QuasiScience developed an easy-to-use pricing model that enabled our client to adjust prices as conditions changed. Our model provides:

Integration of macroeconomic data: including inflation, exchange rates, seasonal fluctuations and other factors not captured by off-the-shelf tools
Automatic price optimisation: Practical recommendations that balance attractiveness to customers with revenue goals and stock availability.
Scenario-based options: Simple scenario-based views so the team could model how different pricing choices might affect revenue in different contexts.
Evidence-based decision making: Easy-to-use alerts indicating when and how prices should be updated.

The tool took the guesswork out of pricing optimisation, and enabled our client to navigate an uncertain market with confidence.

Know Your Assets

The tool developed by QuasiScience transformed the team's work. It led to:

4% revenue increases during peak seasons and high-volatility periods.
More consistent pricing decisions, reducing the guesswork that previously led to lost margin or missed sales.
Better alignment between teams helping teams maximise revenue across the organisation.
A stronger long-term pricing foundation, opening the door to broader revenue-management improvements.

And this tool is not only relevant in retail - it has broad applications across every B2C sector, particularly where companies are operating across multiple regions.

Do you want to engineer better margins? Contact us today.

Digitisation at scale

Mon, 20 Sep 2021 00:00:00 GMT

High-volume scanning systems are essential for many types of research, archive work, and compliance. But capturing data is only the beginning. To make them useful, files must be transferred, backed up, renamed, converted, indexed, enriched with metadata and made available to downstream systems.

Our client, Facture, an exciting art research start-up aiming to make art history accessible to all, was spending >80% of their staff time transforming their scans of art books into usable information for their database. They asked our engineers to help.

From Raw Scans to Usable Data

The existing workflows relied on a series of manual handoffs: someone had to identify new files, move them to the right server, create a backup, convert them into the required format and prepare them for the database and website. But as Facture aimed to expand their collection from 100,000 images to >2 million, the system wasn't working.

Drawing on our decades of experience creating state-of-the-art data pipelines for Formula 1 and other high-tech industries, QuasiScience engineered a custom data offloading application that automated Facture's entire process.

Our application can:

Transfer data to servers, automatically moving newly generated files into downstream processing environments.
Create backups, protecting source data before further processing takes place.
Rename and organise files, applying consistent naming conventions automatically.
Convert images, producing the formats required by subsequent systems.
Automatically generate thumbnails, creating lightweight previews for rapid access.
Generate PDFs, converting scanned material into an accessible format.
Extract text from art books, even if it's in multiple languages or uses special characters.
Automatically generate metadata, enriching files with the information needed for search and downstream processing.
Upload processed data, completing the journey from physical scan to accessible digital record.

A System That Runs Itself

QuasiScience's system turned a painstaking set of manual processes that would have required Facture to hire extra staff into a single automated process. It freed up their existing staff's time to focus on more strategic activities, such as searching for rare catalogues and publications to improve their database. And it provided the foundation for us to build Facture's unparalleled hybrid art recognition software.

What would you have time for if your files organised themselves?

Although in this case we used it to transfer data about art books from a scanning robot, this architecture can be applied to any file-based system, whether you need to tidy up your predecessor's messy files at work, be sure you're not missing anything in a compliance exercise, or prepare large amounts of data for training a Machine Learning Model.

What would you be able to do if your files organised themselves? Contact us today.

Reducing emissions

Mon, 20 Sep 2021 00:00:00 GMT

Scope 3 emissions monitoring is a complex regulatory requirement that asks organisations to trace their environmental impact across an extensive and interconnected supply chain — from raw materials and manufacturing through packaging, logistics and the eventual use of products.

For organisations operating across multiple markets, this becomes particularly difficult as different data providers often produce conflicting estimates for the same activity. On top of this, most existing emissions software is designed primarily for reporting, rather than exploring how operational decisions could actually reduce emissions while meeting commercial goals.

Ortus Climate Mitigation, a renewables company with a portfolio spanning three continents, wanted to understand how it could keep its operations within internal emissions targets while making decisions across a complex and geographically distributed business. They turned to QuasiScience.

From Reporting to Reduction

Drawing on our experience integrating complex data on F1 aerodynamics, QuasiScience's engineering team developed a bespoke modelling and optimisation framework that brings together disparate emissions data, whilst embedding Ortus's own business logic throughout the system.

It offers:

Multi-source data integration, pulling Scope 3 emissions information from different sources across the organisation.
Dataset reconciliation, resolving discrepancies between major emissions datasets including EXIOBASE and ICIO.
Embedded business constraints, incorporating the organisation's operational rules and assumptions directly into modelling.
Scalable optimisation, across a large and geographically distributed portfolio.
Decision-ready outputs, translating complex modelling into information that can be used to reduce emissions.

From Sustainability to Success

Thanks in part to this more robust understanding of its emissions exposure, Ortus was able to secure a contract with Excel Energy, the grid operator of the State of Colorado.

How could your supply chain be greener?

Reporting tells you where you are. Optimisation helps you get where you want to go. Get in touch to discover how to change emissions reporting from a frustrating chore into a genuinely impactful process.

Going Green

Mon, 26 Sep 2022 23:00:00 GMT

As Colorado accelerates its transition to 100% renewable energy, it faces one major challenge - reliability. On days when the wind doesn't blow or the sun doesn't shine, the state still needs to maintain a stable supply and respond to fluctuating demands for energy.

Our client, a renewable energy company, wanted Colorado to invest in their new energy storage facility, but the state was put off by the significant upfront cost.

QuasiScience modelled demand for energy storage throughout Colorado's green transition and proved that flexible and reliable storage solutions will be operationally essential. Our client was then able to proceed to build a solid business case to secure financing.

The high cost of going green

Despite their environmental benefits, power sources such as wind and solar are by their nature intermittent, which introduces major challenges for national grids, who need to maintain a stable supply and respond to fluctuating demands for energy.

This challenge makes energy storage and emergency power control crucial capabilities. Our client, an innovative renewable energy company, wanted to secure investment for a hydropower reservoir, which uses excess energy to store water at height so that its potential energy can be converted into electricity when required.

But these facilities are expensive to build, with upfront costs of >£1bn. Our client needed to prove why this capability would offer a good return on investment.

Numerical simulation de-risks decision-making

Using our proprietary advanced numerical simulation and optimisation techniques, QuasiScience modelled how demand for energy storage changes depending on the uptake of renewable energy. We:

Tested different grid and market scenarios to quantify risk and opportunity.
Optimised operational and investment strategies for different scenarios.
Provided actionable insights for decision-makers, proving the critical role of energy storage as the proportion of renewable energy in the grid increases.

Evidence-based insights enable high-value deals

Thanks to QuasiScience's analysis, the client gained:

Investment readiness: The analysis provided the evidence needed to convince investors of the viability of the project.
Competitive advantage: The insights gained from the simulations allowed Ortus to differentiate themselves in a competitive market.
Long-term planning: The ability to model different scenarios enabled Ortus to plan for the future with greater confidence.

Sustainable Goals

QuasiScience's study reframed Colorado's thinking about hydropower energy storage: this capability is not just a contribution to aspirational sustainability goals, it's strategically essential. Thanks to QuasiScience's analysis, our client was able to demonstrate the strategic value of hydropower storage, linking the capability to measurable reliability gains; as well as to Colorado's clean energy targets.

Numerical simulations and optimisation can de-risk businesses and make investors' decisions easier, and give leaders the confidence they need to make the investments that really matter.

Automated Land Selection

Mon, 26 Sep 2022 23:00:00 GMT

When creating new solar power developments, investors and landowners require precise, site-level assessments to identify viable land parcels. These draw on a wide range of data sources, and require significant expertise to produce, making them expensive and time-consuming.

For our client, Ortus Climate Mitigation, we created an automated, web-based land-suitability platform that merged government perimeter data, ownership records, and environmental indicators using advanced geospatial analytics, to streamline selection of optimal solar-farm locations. This transformed a slow, manual process that took days per site into an automated pipeline that took only minutes for each decision.

A procedural barrier to the green revolution

As Europe accelerates its renewable energy commitments, the demand for utility-scale solar has surged. Yet identifying suitable land parcels remains one of the industry's most operationally complex bottlenecks. Public records are fragmented, cadastral boundaries vary in quality, environmental constraints are scattered across multiple data providers, and manual assessments can take days or even weeks for a single site. The result is a slow, costly, and error-prone process that limits developers' ability to scale portfolios or compete in land negotiations.

Ortus aimed to enter the solar development market with a data-driven edge. Their challenge was clear: evaluate thousands of allotments to determine which were technically viable, commercially attractive, and administratively feasible. This required integrating heterogeneous geospatial datasets, validating boundaries, calculating usable area, and assessing suitability against multiple constraints such as land shape, proximity to grid infrastructure, and ownership status.

The stakes were high: a single unsuitable parcel can cost millions in sunk feasibility work, while a timely, accurate evaluation can unlock early stage rights and investor interest. So Ortus needed a scalable system that translated raw geospatial data into clear development decisions.

An engineer's approach

Drawing on our decades of experience and research in cutting-edge automation technologies, QuasiScience built a web-based geospatial decision platform that automated land analysis to identify high-potential solar farm locations. The approach combined rigorous data engineering with sound business logic to transform messy public records into actionable insights, including:

Automated land suitability scoring: government allotment perimeters, ownership records, and environmental attributes integrated into a unified spatial model, enabling rapid filtering of viable parcels.
Flexible data exploration: an interactive map interface where users could visually inspect parcels, compare attributes, and export candidate sites for further feasibility work.
GIS-native processing: quick and accurate processing of geographical data at scale.

The results

>99% speed improvement: We automated a process that previously required days per parcel to take minutes, enabling rapid, scalable portfolio-wide screening.
A single source of truth: By unifying perimeter data, ownership records, and suitability factors, our system replaced multiple disjointed workflows.
Investment-grade transparency: The platform provided defensible, traceable logic for land selection, improving confidence for investors, regulators, and internal stakeholders, and enabling our client to secure investment in their 1.2GW portfolio of solar farms.
Scalable foundations for future expansion: The data architecture and GIS engine were built to accommodate national-scale land datasets, positioning our client for aggressive market growth.

"You guys rock" -- Remy Martin, CFO, Ortus Climate Mitigation

Are you working in a land-intensive industry, and struggling to manage uncertainty in your investments? Contact us today to get started with faster, more data-driven decision-making.

Project Finance

Sat, 15 Mar 2025 00:00:00 GMT

Renewable assets are increasingly central to forward-thinking infrastructure investors. But pricing is more of a challenge for unreliable renewables than for traditional energy sources. Variability in weather patterns, regulatory changes, and market dynamics all contribute to the complexity of valuing these assets.

Our client, Ortus Climate Mitigation a new energy company investing in a portfolio of renewables projects, turned to QuasiScience for a bespoke simulation system to estimate returns on renewable assets. Our model helped them value accurately a diverse portfolio or renewable energy projects, in preparation for a major acquisition as they expanded their market presence across Europe.

You guys rock!

— Remy Marino, CFO, Ortus Climate

The value of the future

In today's rapidly changing energy market, renewable assets are increasingly central to forward-thinking energy investors' and producers' portfolios. But with opportunity comes complexity. Renewable energy, whether it is wind, solar or hydroelectric, is subject to fluctuating production patterns, shifting market conditions, and (sometimes rapidly) shifting policy frameworks.

The business pages are littered with examples of failed, high-profile renewables projects: from BP's $1.1bn write-down of offshore wind projects to the failure of Saudi Arabia's $200bn solar facility.

Our client was investing in a new portfolio of renewables installations, and wanted a reliable way to predict their financial performance and set prices. They had decades of experience in the energy sector, but their existing tools lacked the flexibility to handle the complexity that comes with renewable energy.

An engineer's approach

Our client needed a model that would take account of:

Intermittent production: Solar output fluctuates with cloud cover, time of day and seasonal changes. Wind power varies hourly and seasonally.
Shifting market conditions: Prices for electricity are influenced by demand spikes, fuel costs, and regulatory changes.
Portfolio effects: Interactions between assets—such as wind farms in different regions—can amplify or dampen risks.

Using advanced Monte Carlo simulations combined with our 200+ proprietary model approaches, we build a reliable evidence-based system to predict returns on different installations, taking account of these factors. Then, we added:

A custom workflow that translated outputs into the client's existing systems in an intuitive, user-friendly way
An automatic bid-price optimisation tool in which staff had only to set the level of desired risk (e.g. 95% probability of returns exceeding a threshold) to receive the optimum bid price for competitive power-provision tenders, balancing risk and competitiveness.

As ever, it's not just about the tech - integrations, training, governance and security are the key to turn smart maths into useful tools for businesses.

The tech behind smart decisions

Using our system, the Ortus team:

Were able to support their expansion into new markets across Europe
Were able to make business decisions based on evidence and share a common language across teams
Began construction of a pipeline of 1.2 GW of solar projects and 1 GW of wind projects

In today's fast-moving business environment, uncertainty is often the only certainty. Traditional statistical methods fail to capture the full range of possible outcomes, and can't cope with extreme scenarios. But advanced simulations like the ones we used in this project enable business leaders to quantify risk with confidence, optimise decision-making, and plan effectively for extreme scenarios. They are the key to navigating uncertainty with confidence.

Do you have a complex investment or business strategy decision you'd like to model? Contact us today.

Recommendation Engines in Retail

Sat, 31 Jan 2026 00:00:00 GMT

As retailers face increasing competition, online recommendation quality has become a major driver of conversion and customer lifetime value. Our client, a S&P 500 listed fashion retailer, asked QuasiScience to improve their systems. We developed a suite of advanced, context-rich recommendation models, generating 3% uplift in total sales, worth several million dollars.

Consumers are expecting personalised experiences; they expect that [we] know who they are — not just that we recognise them when they are online, but wherever they are interacting with the brand.

— Mary Beth Laughton, EVP of U.S. Omnichannel Retail at Sephora

As retailers compete in an environment where customer expectations for personalisation are higher than ever, even small improvements in relevance can translate into large revenue gains. Digital channels are a key arena in which to influence purchase decisions, so recommendation systems that surface the most compelling products, improve conversion rates, and maximise cart value, are crucial to success.

Our client, a major retail brand, was using simple recommendation rules (such as 'most bought items') that failed to incorporate data on user behavior. This meant they could only offer static, low-context suggestions that neither reflected individual preferences nor adapted to the shopper's intent. Our client knew that millions in potential revenue were being wasted, but lacked the modeling infrastructure to capture this opportunity.

True Personalisation

QuasiScience designed a comprehensive, multi-layer recommendation framework that captured far richer context around each customer and their product interactions. Key components included:

Context-Rich Models: Drawing on transaction data, location data, and records of users activities to understand both long-term preferences and immediate intent.
User and product-based recommendations: Tailored 'you might also like' suggestions, as well as complementary and similar item recommendations.
Dynamic experience optimisation: so teams could test and deploy different recommendation strategies and measure their impact in real time.

We implemented these models using scalable pipelines that fit seamlessly into the client's existing digital infrastructure, minimising the amount of training and adaptation required for staff.

Wins for both customers and the business

The project delivered significant commercial value:

A 3% uplift in sales across various categories, totalling an additional £2m over the first year of the pilot.
Faster, evidence-based decision-making, as merchandising, digital, and marketing teams now have clear evidence on which types of recommendations perform best.

Applications Across Industries

Recommendations are not only relevant to the world of fashion - they are an essential tool for any large-scale B2C business. And they also extend beyond the online retail environment - similar techniques can be used to offer real-time in-store personalisation, audience-level targeting, and next-generation omnichannel experiences. However you use them, one thing is clear: cutting-edge recommendation engines are crucial for competitive advantage in an increasingly data-driven retail landscape.

Slash hosting costs

Mon, 20 Sep 2021 00:00:00 GMT

Cloud hosting has transformed how businesses build and operate digital products, but without careful management, costs can grow just as quickly as the infrastructure itself. As organisations scale, introduce new services, and process larger volumes of data, cloud expenditure often becomes one of their largest operational costs.

By understanding and optimising cloud hosting costs, businesses can improve profitability, allocate resources more effectively, and ensure their technology infrastructure scales sustainably alongside their growth.

Our outcome-based introductory services

Our expert engineers work closely with world-class Universities to develop novel cloud hosting technologies. So sorting out your hosting costs is easy. For new clients, we offer an outcome-based cloud compute optimisation service - if we don't save you money, you don't pay.

If you're making a significant investment in cloud computing, and you're not sure if it's paying its way contact us today.

Smart repairs

Mon, 20 Sep 2021 00:00:00 GMT

Solar plants generate a continuous stream of operational data, but identifying when a plant is underperforming is not that simple. The challenge is to distinguish genuine faults from normal variation in data and to understand when maintenance is actually required.

In 2019, QuasiScience supported Ortus Climate Mitigation by developing a digital twin for its solar plants - an advanced simulation capable of tracking live plant outputs, comparing them to a mathematical prediction of performance, and flagging anomalies.

Maintenance is a major cost

Traditional maintenance strategies usually require engineers to inspect equipment on a fixed schedule, regardless of whether intervention is actually necessary. For renewable assets spread over a large area, this can create significant unnecessary costs.

Our team of award-winning engineers developed a sophisticated digital representation of normal solar plant behaviour, that then continuously compared expected and observed outputs to identify potential problems. The system provides:

Continuous performance monitoring, tracking outputs over time.
Anomaly detection, identifying deviations from expected behaviour.
Actionable maintenance information, helping teams determine when intervention is actually required.

From Scheduled to Just-in-Time

Our system gave Ortus a more intelligent basis for deciding when maintenance was necessary. The approach enabled a just-in-time maintenance strategy, reducing unnecessary costs.

Can your assets tell you when they need attention?

Intelligent monitoring systems, or digital twins, like this one turn operational data into actionable information, helping organisations maximise efficiency and minimise costs.

If you're operating a business with a large portfolio of physical assets and you wish they could tell you what was wrong, talk to us today.

21st century security

Mon, 20 Sep 2021 00:00:00 GMT

As technology companies grow, their infrastructure and development practices often evolve faster than their security controls. Misconfigurations, inconsistent permissions and limited visibility can create vulnerabilities that become increasingly difficult to address as systems scale.

A leading technology scale-up we were working with on another project asked QuasiScience to prepare its organisation for a cybersecurity compliance audit. The objective was not simply to fix individual vulnerabilities, but to establish sustainable security practices that could continue as the company and its technology estate grew.

From Messy Fixes to Security by Design

QuasiScience conducted a review of the company's infrastructure and identified several major vulnerabilities. Rather than treating these as isolated technical problems, our F1-trained team helped establish a longer-term approach to security management across development and production environments.

Our work included:

Azure security remediation, identifying and fixing misconfigurations.
Kubernetes security policies, establishing appropriate controls across development and production clusters.
Vulnerability scanning, introducing systematic identification of security weaknesses.
Observability improvements, improving visibility into infrastructure and application behaviour.
Role-based access control, ensuring permissions were appropriately managed across systems.
Sustainable security practices, embedding cybersecurity into the company's ongoing technology delivery processes.

Ready for the Audit

The work transformed the company's cybersecurity approach from reactive fixes into a more systematic management process. Based on our work, the company achieved the SOC 2 certification it needed for the next stage of its growth.

Is your security keeping pace with your growth?

In the AI-era, cybersecurity is more complex and evolving quicker than ever. Therefore, compliance should not be treated as a one-off exercise before an audit. The strongest security programmes build the right controls into the way technology is developed and operated every day. Based on our ongoing research and deep familiarity with emerging technologies, QuasiScience helps growing technology businesses identify vulnerabilities, strengthen their infrastructure and build security practices that scale with them.

Social travel

Mon, 20 Sep 2021 00:00:00 GMT

In 2026, we're not short of travel platforms to help us search for hotels, flights or experiences. But what they're less good at is finding the right combination of options.

Our client, Tribyou wanted to build a different kind of travel platform: one that could bring together accommodation, destinations, experiences, services, and even fellow travellers. Their platform needed to do more than display listings. It needed technology capable of understanding a user's dream holiday, helping them find their 'tribe', and plan their ideal trip.

More Than a Booking Website

The Tribyou platform will eventually host thousands of possible destinations, fellow travellers, experiences and services. Combined with each user's particular preferences, finding the trip they will enjoy most becomes an optimisation problem.

Drawing on our 200+ proprietary code libraries, our team engineered the platform to provide a secure infrastructure for mathematical simulation and optimisation.

Tribyou's platform can now:

Optimise recommendations, identifying combinations of options that best satisfy users' preferences.
Balance competing requirements, accounting for the fact that different users value different things.
Apply hard constraints, ensuring that non-negotiable requirements are always respected.
Incorporate feedback, allowing matching criteria to evolve as the platform learns more about what users like.
Connect the optimisation engine to the wider platform, allowing the underlying mathematics to operate as part of a real customer-facing product.

From Search to Service

In the competitive world of online retail, users expect personalisation. The best websites aren't just websites - they efficiently match users with the products or experiences they are most likely to value. That's only possible with mathematical optimisation, not just manual curation.

If you want to offer your customers more of what they really want, contact us today.

Specialist cloud system

Sun, 31 Mar 2024 00:00:00 GMT

Sound Mathematics is an award-winning UK startup that specialises in non-destructive testing. While their core technology, a Machine Learning algorithm producing full health reports for metallic components, has won numerous awards, they were not set up to deliver their insights as a service. They reached out to QuasiScience to help them build robust and cost efficient software architecture to start serving their first clients.

Challenges

After compiling a full list of requirements for the architecture, QuasiScience and Sound Mathematics agreed that the key requirements were:

Users of the service are required to upload large files.
The service should return a complete analysis in a matter of seconds.
The cost at rest should be as close as possible to zero without impacting the potential to scale.
Project duration: less than a month to avoid impacting commercial partners.

Implementation

Serverless

QuasiScience designed and built a serverless architecture to store, process, and return predictions on a large number of files. The architecture is based on AWS Lambda, S3 and DynamoDB. This allows it to scale horizontally so our clients only pay for the resources they use.

Containerisation

QuasiScience containerised the inference code using custom built OS images to reduce the memory footprint by ensuring only the necessary libraries and dependencies are included. This allows us to easily update the code and scale the service without worrying about dependencies or compatibility issues.

Multistage Processing

Users are expected to upload large files to the backend together with special configurations. This process can be slow for the user and costly for the server provider. For this reason, we split the inference process atomically to catch user errors as quickly as possible and allow users to break up the data upload and result collection process.

Results

Thanks to the joint effort of our teams, Sound Mathematics was able to quickly integrate the APIs QuasiScience built for them in their frontend and cloud architecture to engage with their customers and keep moving forward with commercialisation.

Quality architecture is often the difference between a great idea and a real-world product.

Streamlined Materials Performance Testing

Sat, 31 Jan 2026 00:00:00 GMT

QuasiScience helped a leading advanced materials company transform how it manages and uses its testing data. Drawing on our experience building data systems for Formula One teams, we developed a unified materials data platform that centralised experimental results, automated data capture, and enabled more effective analysis and collaboration. This accelerated R&D, improved material selection, and preserved valuable institutional knowledge.

Effective data management can be the key to improving efficiency in materials research and opening up its added value.

— Dr Ben Thomas, Department of Materials Science & Engineering, University of Sheffield

Materials Testing

Industries from aerospace to automotive run hundreds of materials testing experiments every year. For any business involved in producing physical products, it's crucial to understand how different materials, or combinations of materials, surface treatments and adhesives perform under varying conditions.

However, many organisations are not making the most of this data. Results may be stored in disjointed spreadsheets, lab notebooks, or historic databases. This fragmented approach is far from the best foundation for research and decision-making.

Our client, a leading advanced materials company, wanted to unlock the power of its data by improving the way it collected, organised and analysed the results of experiments, past and present.

From F1 to the Factory

Drawing on our team's experience building high-performance data systems for Formula One teams, QuasiScience designed and implemented a unified data platform tailored to the client's needs.

When we worked in F1, we needed to track variables such as composite layering, surface coatings, adhesive types, and mechanical test outcomes, to select the optimal combination for each component, balancing strength, weight, and durability. We built a similar system, featuring:

Centralised Data Collection: Automated data capture from multiple lab instruments and testing set-ups.
Smart Organisation and visualisation: to make experiments easily searchable and comparable.
Custom analytical tools: Built-in visualisation tools to analyse trends, correlations, and trade-offs using custom metrics.
Collaboration Features: Live, secure, role-based access for engineers, researchers, and management.

Information management in R&D

QuasiScience's new data platform transformed the client's R&D process. For the first time, engineers and scientists could view all test results in one place, compare outcomes across experiments, and quickly identify promising combinations. This meant:

Faster innovation: Engineers reduced the time spent searching for past data and setting up redundant experiments.
Improved decision-making: data-driven selection of materials based on all the evidence available.
Knowledge retention: Institutional knowledge from years of testing became easily accessible, reducing reliance on individual experts
Enhanced collaboration: Multiple teams across different sites could access consistent, validated data, accelerating joint research.

The Good Data Advantage

Data infrastructure may sound like a dry topic, but without good databases and working data pipelines there is no methodology that could yield a good outcome. What is the use of advanced simulations, machine learning models or digital twins if the underlying data is poor quality, incomplete or inaccessible?

Good infrastructure makes sure data is clean (free of structural issues like typos, duplicates and missing entries); accurate; legally usable; secure and accessible, and opens up the opportunity for companies to use all the latest data science and research techniques.

Taming data

Mon, 20 Sep 2021 00:00:00 GMT

It goes without saying that there's no data science without data. But there can be too much of a good thing: as organisations scale, they often accumulate datasets across projects, experiments and systems; and knowing what exists, and where to find it, can become more difficult than the analysis itself.

Our client, an ambitious scale-up in the safety industry, had accumulated more than 10TB of datasets, but they were spread across a messy file structure. Their data science team needed a way to discover what was available, understand how datasets were organised and identify gaps in the information required for their work.

From chaos to capability

In our experience, large datasets are rarely organised in a neat database. They evolve over time, reflecting the history of projects, teams and experiments. It's not realistic to expect busy teams to adopt perfect information management practices, and in any case, a single file structure may not suit every type of data the organisation holds.

For our client, we built a custom internal portal designed around their existing data architecture. Rather than forcing the organisation into a time-consuming and boring reorganisation of its datasets, our world-class engineers built a platform to understand and navigate the structure that was already there.

The platform provides:

Discovery, allowing data scientists to navigate datasets smoothly.
Visibility, making fragmented information easier to find across more than 10TB of data.
Gap identification, helping teams understand what data exists and what is missing.
Security, preparing the organisation to meet industry standards as it scales.
An extensible foundation, allowing the platform to evolve as datasets and analytical requirements grow.

How much valuable data is hiding in your organisation?

You've probably got a lot of data. But that's not enough to turn your business into a science. If your teams cannot find it, understand it or identify what is missing, data remains without value.

If you want to make your data usable, contact QuasiScience today.

The perfect host

Mon, 20 Sep 2021 00:00:00 GMT

In 2026, matching-based service businesses are everywhere - from dating to hospitality to recruitment, matching users with a provider or fellow user, rather than providing a service directly, can enable businesses to be more responsive to customer needs, whilst reducing the resources required to operate at scale.

At first, matching sounds simple. But in practice, it's complex - every user has different requirements, preferences and constraints, and providing the best outcome for one person may make it harder to find a good outcome for someone else.

TheIcircle helps students find the perfect host for their time studying abroad. They asked QuasiScience to solve exactly this problem: how to match hundreds of students with suitable host families while accounting for the different requirements of both sides of the relationship, and the unique safeguarding requirements of a platform for students.

The Right Match for Everyone

Drawing on our state-of-the-art knowledge of F1-standard optimisation algorithms, QuasiScience engineered a bespoke system, which:

Matches students and hosts, balancing sometimes-competing requirements.
Incorporates user feedback, so match requirements evolve based on experience.
Applies hard constraints, ensuring safety and business requirements are respected.
Optimises the overall allocation, maximising average match quality rather than simply producing the best individual matches.
Operates at scale, matching hundreds of students and hosts efficiently.

Life-Changing Opportunities

QuasiScience's optimisation system allows theIcircle to match students with safe, welcoming host families, and hosts with reliable, respectful students. Our secure, scalable software is providing a foundation for global expansion, creating educational opportunities for hundreds of students.

Who do you need to match?

Optimising customer matches with services, products or fellow users can have massive impact on businesses across sectors. But only if its done right. Unsatisfactory matches can alienate customers, and damage brand value.

For matching you can trust, contact QuasiScience today.

Therapy platform

Sun, 31 Mar 2024 00:00:00 GMT

Customer experience isn't just about profit

Mental health: there is no sector where customer experience is more important. Any friction can erode trust and discourage individuals who are already in vulnerable situations.

Our client, Coyzy, an Italian mental health company, understands better than most the importance of providing first class customer experience. However, as their platform evolved across multiple devices and service tiers, they were facing growing technical challenges that were affecting user experience. They had to manage different user identities across platforms built at different times and using different backend technologies; and combine data from third-party integrations (such as video call and scheduling tools) with their internal records for billing, but this was largely done manually, leaving room for error.

The result: customers were having trouble accessing services, and sometimes even receiving incorrect bills - another unnecessary stressor for people already seeking support with their mental health.

Engineering mental health

Coyzy reached out to QuasiScience to design and implement a scalable, integrated system that would not only address their immediate challenges, but also lay the groundwork for future growth.

1. Intuitive identity management

Drawing on some of our 200+ proprietary code libraries, our team of expert engineers developed a robust but efficient identity management framework capable of supporting logins from multiple backend architectures. By integrating authentication protocols and creating a centralised user database, we ensured Coyzy's customers could sign in effortlessly from any device, but still expect best-in-class security. Coyzy's core development team also gained a single source of truth for user profiles and permissions.

2. Automated appointment tracking

We built an appointment tracking and reconciliation system that automatically records, verifies, and matches user sessions with billing data.

This eliminated manual reconciliation, which had a significant impact on Coyzy's administrative overheads, and improved accuracy.

3. Automated billing

Working closely with Coyzy's finance and technical teams, we streamlined the billing process by linking subscription data with real-time usage. This enabled the company to automate billing across its tiered pricing model, improving efficiency, customer experience and safeguarding.

The new systems brought immediate benefits. Login issues dropped significantly; user engagement improved; the automation of billing and appointment reconciliation freed up valuable staff time; and, most importantly, Coyzy gained a scalable digital foundation to support future product development and market expansion.

We don't see QuasiScience as a supplier, but as part of the team. You're a precious resource - reliable and competent.

— Luca Piras, Co-founder and CEO, Coyzy

Looking ahead

With a solid technical foundation in place, Coyzy is well-positioned to continue its mission of promoting mental well-being. The company plans to use its enhanced platform to introduce new features, expand its user base, and explore partnerships that further its impact in the mental health space.

QuasiScience remains a trusted partner, ready to support Coyzy as it navigates the evolving landscape of digital mental health services.

Do you need to build or improve a secure, automated user platform? Contact us today.

Ultra-fast drone

Mon, 20 Sep 2021 00:00:00 GMT

When performance is a matter of life or death, small engineering decisions can't be left to intuition. Working with a high-tech Ukrainian start up, QuasiScience helped develop a high-performance drone, optimising engine positioning to achieve maximum performance.

Optimising Every Millimetre

In F1 cars or in drones, engines need to be positioned where they can deliver power most effectively, but where their location will have minimal negative impact on aerodynamic drag around the vehicle.

QuasiScience analysed:

Aerodynamic drag, assessing how different engine positions affected resistance.
Power efficiency, ensuring that engine placement did not unnecessarily waste power.
System-level effects, balancing individual constraints against the performance of the complete drone.

Rather than relying solely on trial and error, the team in Ukraine were thus able to explore design options scientifically and identify a model that balanced competing requirements.

Record-breaking results

The analysis enabled our client to produce a drone designed for extreme speed - at the time, the fastest in the world. From Formula 1 to the front line, mathematical modelling is essential for effective engineering.

AI-ready images

Mon, 20 Sep 2021 00:00:00 GMT

AI has been a boon for film editors, removing much of the painstaking manual labour that used to take months or years. But modern machine learning systems require carefully structured training data. When it comes to editing films, that means raw video and image collections need to be turned into precisely formatted datasets: frames must be extracted, formats must be changed, images need to be cropped, filters or colour masks need to be applied, and visual data needs to be separated into the components required for training.

Our client, Nulight studios, a boutique film production company, needed a way to get their visual data ready to be edited with the latest machine-learning powered technologies, producing work that would otherwise only be possible for a big-budget commercial studio. They asked QuasiScience to engineer their data preparation pipeline.

From Manual Processing to Automated Pipelines

QuasiScience built an automated media-processing pipeline around FFmpeg, a high-performance library for working with video and audio. The system allows users to describe the transformations they need in a small number of natural language instructions, then automatically applies those operations across the entire source dataset.

The pipeline can perform:

Format conversion, ready for downstream processing.
Cropping and resizing, so visual data becomes suitable for specific machine learning architectures.
Frame and image extraction, splitting video into individual images or other suitable training-data components.
Colour transformations and filter passes, applying consistent colour changes across large datasets.
Batch processing, executing complex sequences of transformations across entire films.
Comprehensive machine learning preparation, producing consistent, structured datasets ready for model training.

Turning Films into Training Data

The resulting system transformed visual-data preparation from a painstaking task that took as long as manual editing, into an automated, repeatable process. Building on this work, we worked with Nulight to develop a groundbreaking Machine Learning powered object removal software, which won the 2024 award for Amplifying Imagination: AI in the Creative Industries, sponsored by the BBC, AWS, Innovate UK and the Digital Catapult.

How could your data be better engineered?

If you want to use machine learning tools on video or images, or if you need to process image or video at scale, manual data preparation quickly becomes a bottleneck.

Drawing on our 200+ proprietary code libraries and our award-winning search, QuasiScience can help you build robust data pipelines that turn complex requirements into simple, reliable workflows. Get in touch to discover what your data can really do.