Synergy

Finding the right co-founder is one of the hardest challenges in entrepreneurship. We wanted to build a platform that could understand what founders are looking for through natural conversation, then intelligently match them with complementary teammates based on skills, values, and needs rather than just surface-level criteria.

What:

Synergy is an AI co-founder matching platform that engages founders through guided conversation and extracts structured profiles to find their perfect teammates. Users interact through a dual-interface approach: a freeform prompt or a persistent chatbot that naturally collects information about background, skills, and collaboration preferences. The platform then uses semantic search and intelligent matching to recommend complementary co-founders, with the UI visualizing compatibility scores and team suggestions.

Build/Tech Framework:

We paired a polished Next.js 14 frontend with a Flask-based AI backend. The frontend uses shadcn/ui, Tailwind CSS, react-chatbot-kit, and Lucide icons for a high-end conversational interface. The Flask backend leverages Groq's Llama-3 models: an 8B instant model for rapid chat and a 70B JSON-mode model for profile extraction. We generate embeddings locally with SentenceTransformers and persist them into ChromaDB, creating a semantic knowledge base that powers our matching algorithm.

Project Challenges:

Building intelligent matchmaking proved complex. We had to guide users through structured conversation without feeling robotic, extract clean JSON profiles from freeform dialogue, and move beyond simple vector similarity to recommend genuinely complementary teams. Balancing speed with quality by using the faster 8B model for chat and the 70B model for profile extraction also required careful engineering.

Accomplishments:

We built a system that feels genuinely intelligent rather than a generic GPT wrapper. Our matchmaking layer normalizes roles, applies semantic filters, and re-ranks candidates based on team composition. The adaptive questioning naturally collects needed information, and watching conversational data transform into structured profiles and meaningful team recommendations is satisfying. The UI successfully makes the entire process transparent and easy to understand.

What we learned:

We learned that AI systems need discipline and structure to be useful. A good language model isn't enough; you need conversational policies, validation, and careful post-processing of outputs. Building a good UX around AI is hard, managing state across multiple entry points requires careful thought, and combining multiple models with different strengths creates better outcomes than relying on a single approach.

What's next for Synergy?

Well, we're exploring richer founder profiles capturing working styles and values, real-time availability filtering, feedback loops to improve matches over time, and team formation workflows for newly matched founders. We're also investigating explainability features so founders understand not just who they're matched with, but why.