Multi-Agent Passion Intelligence

Early-stage startups lose the hiring game on the terms Big Tech sets — salary, brand, and a recruiting team they don't have. But there's one axis they can win on: finding the few people already, voluntarily, building toward their mission. Resumes describe what people say they can do. Side projects reveal what they can't stop building. Multi-Agent Passion Intelligence turns that signal into a 10-agent LangGraph investigation, powered by Gemma on an AMD Instinct MI300X. You give it your mission; it does the rest. 🔭 FREE DISCOVERY — you supply no names. It searches GitHub from your mission alone and surfaces builders who never applied to you, and never would have found you. Sourcing, not screening — the part no other hiring tool does. 🔍 EVIDENCE, NOT VIBES. It reads their GitHub repos, hackathon projects, and writing across lablab.ai, Dev.to, Hacker News, Devpost, and Kaggle — and Gemma's vision model reads their architecture diagrams and product screenshots. Every score cites a real URL, so a founder can verify any claim in one click. Nothing is invented. 📊 RANKED AND REACHABLE. A transparent weighted score — project fit, genuine passion, evidence quality — ranks them, and a candidate is only marked a match if there's a LinkedIn to actually reach them on. A map shows where they are. 🗂 FOUND ONCE, KEPT FOREVER. Everyone it discovers is saved to a reusable talent pool — so the next search surfaces new builders instead of re-listing the ones you already have. 🎬 A PITCH YOU CAN FORWARD. It renders a narrated recommendation video, captioned by Gemma for two audiences: a technical co-founder who wants the stack and the architecture, and a non-technical partner, recruiter, or investor who wants the story. For a startup, this replaces the sourcer you can't afford and the recruiter seat you can't justify. It runs on open Gemma — no per-seat SaaS tax — and it turns "we can't compete for talent" into "we found the three people who were already building this."