Inspiration
Our inspiration came from observing the growing challenges business owners face in finding the right accelerator programs tailored to their specific needs. We wanted to streamline this process by creating a recommendation engine that not only matches businesses with accelerators but also empowers even those with little to no data to find the right growth opportunities. This sparked the idea to integrate AI-driven solutions like vector search and computer vision for a more interactive, user-friendly experience.
What it does
The project consists of an intelligent recommendation engine that uses a vector search algorithm to analyze a company’s products, industry, and entitlements, matching them with accelerators that best fit their needs. We also added a computer vision-powered feature where users can dynamically create display boxes using hand gestures, with OpenAI’s Whisper API transcribing spoken words into text within these boxes. This seamless combination of voice input and gesture control creates an intuitive and highly interactive user experience.
How we built it
We built the core recommendation engine using a highly optimized vector search algorithm and React.JS, that transforms business attributes into high-dimensional vectors. This allowed us to efficiently match businesses with relevant accelerators, even without historical data. For the interactive component, we integrated Mediapipe’s computer vision for hand gesture recognition, alongside OpenAI’s Whisper API for real-time voice transcription. We tied all the elements together into a cohesive system using Python, Flask, and other supporting libraries.
Challenges we ran into
One of the biggest challenges was ensuring that our vector search algorithm could handle companies with little or no data. We also faced difficulties integrating Whisper’s voice transcription with real-time gesture recognition to create a seamless experience between spoken input and visual feedback. Moreover, maintaining low latency while processing both the voice and gesture data in sync was another technical hurdle we overcame.
Accomplishments that we're proud of
We’re particularly proud of how we combined different AI technologies—vector search, computer vision, voice transcription and generative AI—into one smooth and user-friendly platform. Building an accelerator recommendation engine that works for businesses of all sizes, regardless of their historical data, is another milestone we’re proud to have achieved.
What we learned
We learned how powerful combining different AI technologies can be in creating innovative, user-friendly solutions. Integrating vector-based search with gesture-based interactions and real-time voice transcription opened up new possibilities for how users can engage with digital tools. We also deepened our understanding of the complexities of real-time AI processing and synchronization.
What's next for SprintScout
Looking forward, we plan to expand our recommendation engine by incorporating additional data sources to further improve accuracy. We also aim to refine the AI-powered accelerator, adding more interaction methods and enhancing the overall experience for business owners. Ultimately, we see SprintScout evolving into a comprehensive platform where businesses can seamlessly connect with the right resources for growth.
Log in or sign up for Devpost to join the conversation.