The snapshot
1
Empire Games, a London-based gaming studio, needed to automate the creation of playable levels for their mobile puzzle games to reduce manual effort.
2
The challenge was to create a scalable and reliable pipeline that could algorithmically generate valid puzzles from defined parameters with minimal human intervention.
3
The solution involved a new Python-based, LLM-driven workflow on AWS, which used an agentic approach and validated each generated gameboard against a specific schema.
The Customer and Context
Empire Games is a London-based gaming studio focused on creating high-quality, competitive mobile games that offer players both fun and the chance to win real cash. One of their key titles is Nuts & Bolts Screw Wood Puzzle, a competitive, skill-based single player puzzle game. they released Empire Bingo, a rich, narrative-driven bingo experience set across varied thematic backdrops like Ancient Egypt and Victorian England.
Devoteam in collaboration with AWS, created a Proof of Concept to show the feasibility of streamlining Empire Games’ level creation process using GenAI and cloud-based technologies.
The Challenge
The team needed core logic to algorithmically generate new playable levels from defined input parameters and to orchestrate gameboard generation. The pipeline had to be scalable, fault-tolerant, reusable, reliably interpret prompts to produce valid puzzles with minimal manual input, and be highly automated and low-friction for developers and end users.
The Solution
The solution centred on building a Proof of Concept (PoC) for a fully LLM-driven workflow that could handle every stage of gameboard generation from start to finish. At the outset, Claude thinking tokens were used to interpret the initial input and guide subsequent requests, ensuring the logic of each gameboard aligned with the intended puzzle design. Tool integrations were then applied to execute specific operations with precision, reducing the risk of errors or inconsistencies.
Every gameboard produced was validated against a defined schema before delivery, providing a safeguard for quality and playability. To maintain coherence throughout the process, an agentic approach was adopted, preserving context across iterations and allowing the system to generate consistent, reliable outputs at scale.
The chosen methodology was a Python-based orchestration integrated with Amazon Bedrock, with schema validation and an agentic workflow to preserve context throughout generation.
Main Results and Benefits for Empire Games
The Proof of Concept successfully demonstrated that LLM’s and AWS Cloud technologies could automate playable level generation in a way that had not been practical before.
1
LLMs were able to assist gameboard creation, validating the PoC approach.
2
Automated generation was shown to be more cost-effective and scalable than manual methods.
3
The project identified areas for improvement in pattern consistency to further enhance quality and reliability.
The solution was successful because it allowed us to automate puzzle generation in a way that had not been practical before. Achieving a 70% rate of usable gameboards out of 10 generated during our initial proof of concept demonstrated both feasibility and progress towards an efficient solution. We were shown that we would reduce the time of level creation compared to manual methods, while maintaining consistency.
What we valued most about working with Devoteam was their ability to design a Proof of Concept that was technically sound. Their added value was in bringing structure and automation, with potential for large scale to a process that was fully manual and as such, slower.
For other companies facing similar challenges, my advice would be to focus on building a clear validation pipeline early. Automating content generation is only useful if the outputs can be trusted. Partnering with a team like Devoteam that understands both the cloud infrastructure and AI workflow design made that possible for us.
Alex Palaghita
Founder & CTO,
Empire Games
This project exemplifies the transformative potential of Generative AI on AWS for game development. Working with Devoteam, we were able to create an innovative, LLM-driven workflow that significantly streamlined Empire Games’ level creation process.
The impressive 70% success rate achieved in the initial proof of concept demonstrates both the power of our technology and the expertise Devoteam brought to the table. We’re thrilled to see how this solution has positively impacted Empire Games’ efficiency and look forward to supporting their continued innovation in the gaming industry.
Nathan Merola
Enterprise Account Manager,
AWS
Your Success Starts Here
Partner with Devoteam to leverage award-winning tech expertise, agile execution, and a culture of continuous learning.
