Skip to content
← All work

Team project · AI

Lynx

A RAG co-pilot that turns technical IP into an investor thesis, then pitches it to a simulated VC.

Year
2026
Role
Team project
Stack
Python, FastAPI, RAG, Gemini, ElevenLabs, React
Links
Visit site
Source
Lynx landing page: the word LYNX in glowing green outline letters on black.

The problem

Technical founders sit on pages of IP documentation that investors won't read. Turning it into a thesis, and working out which investors are actually a fit, takes hours.

What I built

A FastAPI RAG pipeline parses the documents into a structured knowledge graph. A REST layer connects that to an investor-matching engine, and a real-time pitch simulator uses Gemini for investor personas and ElevenLabs for their voices.

A decision worth explaining

Async end to end, so the simulated VC can keep up

Every turn of the pitch simulator calls both Gemini and ElevenLabs. Handling those requests asynchronously kept responses under 800 ms, fast enough for the exchange to feel like a conversation rather than a queue.

Outcomes

  • Parses 50+ pages of technical documentation in under 2 minutes, down from several hours by hand.
  • 85% mandate-alignment accuracy across 40+ investor profiles.
  • Sub-800 ms responses in the live pitch simulator.

Demo