Crakd
AI-Powered Talent Identification for Developers
Crakd identifies talented developers by combining GitHub metrics with LLM analysis. The system uses ensemble models to score developers based on natural language queries, balancing quantitative code metrics with qualitative profile assessment.
Built at B.E.L.L.E's SF AI hackathon in September 2025.
Architecture
- • Ensemble models combine multiple ML algorithms for developer scoring
- • GitHub API integration for repository metrics and commit analysis
- • Gemini LLM processes natural language queries and profile text
- • Hybrid scoring weighs quantitative metrics with qualitative insights
- • Semantic search enables descriptive developer queries
- • Real-time dashboard for ranking visualization
- • Local analysis tools for detailed profile examination
Tech Stack
- • Frontend: React.js, Vite, Tailwind CSS, Framer Motion, 3js
- • Backend: Python FastAPI, Docker on Render, local analysis using with PCA (matplotlib/scikit-learn)
- • APIs: Google Gemini API, GitHub GraphQL
- • Deployment: Vercel (frontend), Render (backend)
- • Dev tools: Trae, TestSprite, neovim, Gemini CLI
Impact
Moves beyond resume screening by analyzing actual code contributions and GitHub activity. Natural language queries make technical talent search accessible to non-technical recruiters while maintaining analytical depth for accurate assessment.