Founder-AI

Inspiration

Raising VC funding takes months of meetings, pitch decks, and negotiations. We wondered: could AI agents handle the entire process autonomously? Google Gemini's multimodal capabilities made this possible.

What it does

FounderAI is a complete VC platform powered by multi-agent AI:

• Analyzes pitch videos/PDFs with instant scoring (0-100)
• Conducts automated due diligence with verification questions
• Simulates investment committee debates (3 AI personas)
• Negotiates term sheets autonomously with intelligent deal detection
• Generates professional term sheets from conversations
• Predicts future outcomes with crisis scenarios

From pitch to signed term sheet in minutes, not months.

How we built it

Tech Stack:

• React + TypeScript + Vite
• Google Gemini 3.0 Flash (multimodal AI)
• Tailwind CSS for UI
• Structured JSON generation for term sheets

Architecture: Multi-agent agentic AI system where each agent has specific roles:

• Analyzer Agent: Evaluates pitch materials
• DD Agent: Extracts and verifies claims
• Committee Agents: 3 personas debate viability (CTO, CFO, Partner)
• Negotiator Agent: Autonomous deal-making with auto-cancellation
• Term Sheet Generator: Extracts agreed terms from conversations

What we learned

• Gemini's structured JSON output is incredibly powerful for term sheets
• Agentic AI needs clear "exit conditions" (deal completion vs cancellation)
• Free tier API limits require smart optimization (we hit 4 keys in 10 mins!)
• Multi-agent debates create realistic, nuanced evaluations

Challenges we faced

1. API Rate Limits: Free tier = 15 RPM. Solution: Disabled auto-cancel checks to reduce calls by 50%
2. Deal Detection: Too strict = missed deals, too loose = false positives. Balanced with 2+ phrase detection
3. Term Sheet Extraction: AI needed context on full conversation, not just last messages
4. State Management: Negotiation → Success Modal → Term Sheet → Board Sim flow was complex

Built With

• gemini-ai
• react
• tailwindcss
• typescript