Back to Projects
ScoutLayer
AI-first VC sourcing and screening system built solo for a global hackathon.
Next.jsTypeScriptTailwindTavily
The Problem
VC deal sourcing and early screening is slow and manual .. analysts spend hours reading decks and doing basic diligence before a deal is even worth a real look.
The Solution
Built a four-agent pipeline.. Sourcing, Screening, Diligence/Verifier, Decision/Synthesizer using Groq (gpt-oss-120b) and GPT-4.1 Mini for reasoning, Tavily for research, on a Next.js/TypeScript/MongoDB stack with a custom "Terminal Ledger" design system. Solo build, 24-hour hackathon.
// what I learned
Coordinating a four-agent pipeline (Sourcing → Screening → Diligence → Decision) surfaced how much prompt design matters for hand-offs between agents — a vague output from Screening would cascade into bad Diligence calls. Building the "Terminal Ledger" design system also reinforced that a strong visual identity, done fast, makes a solo hackathon build feel far more finished than the code alone would suggest.
