Projects Matcher (The AI-Powered Co-Founder Matchmaker)

💡 Inspiration We were inspired by the common struggle among STEM students: having brilliant project ideas but lacking the necessary complementary skills to execute them—like a coder needing a designer, or a biology student needing a software engineer. Existing networking is slow and keyword-based. Our goal was to build an AI-powered co-founder matchmaker. We embraced the "Mountain Madness" hackathon theme of Jekyll and Hyde lore to give the platform a unique personality.

🛠️ How we built it The platform is built with a React TypeScript frontend using Expo Router to support both web and mobile access. We used ReactBits and Framer Motion to enhance UI/UX with professional animations. The backend uses a Python (FastAPI) framework to interface with our various AI models and database.

For core data and authentication, we utilized Supabase (PostgreSQL), enabling the pgvector extension for semantic matching. We chose Supabase Auth to allow secure Gmail or email logins.

Our AI integrations include:

Google Gemini API to generate intelligent matchmaking based on vector embeddings of user bios and project descriptions.

Gemini API as Dr. Jekyll (The Architect) to convert vague one-sentence project ideas into full technical roadmaps, including DB schemas, API endpoints, and task lists.

ElevenLabs as Mr. Hyde (The Hype-Man), providing chaotic "shitpost" energy by voice-roasting incomplete profiles or hyping exciting projects with high-energy audio commentary.

We prioritized complementary skill matching, domain interests, commitment levels, and experience compatibility rather than just similar job titles.

🚧 Challenges we faced We faced the common hackathon challenge of balancing high ambitions within 24 hours. When our initial concept felt "too bland," we took the technical risk of implementing vector search and complex multi-modal AI integrations simultaneously. Ensuring effective semantic matchmaking required careful prompt engineering and effective vector embedding, especially when using Gemini to generate hypothetical documents to improve matching quality. Managing the latency between AI generation and the UI required creating engaging loading states and asynchronous processes.

🧠 What we learned We gained significant experience in full-stack development using modern web technologies like React, FastAPI, and Supabase. We mastered vector search and semantic matching using the pgvector extension, shifting our perspective on how algorithms can identify "compatibility" over mere similarity. We also learned to design for high-end UI animations with libraries like ReactBits. Deploying on the cloud taught us the importance of persistent backends and secure authentication workflows. Ultimately, we learned how effectively combining text analysis and audio generation can create a compelling "human" user experience that stands out to judges.

Built With

• api
• elevenlabs
• exporouter
• fastapi
• framermotion
• gemini
• google
• pgvector
• postgresql
• python
• react
• reactbits
• supabase
• tts
• typescript