Yalie Search
AI-powered student directory with multimodal search — combining CLIP, semantic text, and GPT-powered query expansion across 6,000+ Yale students with 2-3x retrieval accuracy.
Overview
Yalie Search is an AI-powered student directory enabling natural language queries across 6,000+ Yale students by interests, skills, and appearance — achieving 2-3x retrieval accuracy vs. keyword search through GPT-powered query expansion.
Architecture
Multimodal search architecture combining three complementary systems:
- CLIP Vision Embeddings — Profile images encoded with CLIP ViT-Large-Patch14 for visual similarity
- Semantic Text Search — Natural language understanding for interest and skill-based queries
- Structured Metadata Matching — Filters for residential college, graduation year, and major
- GPT-Powered Query Expansion — Intelligent reformulation for broader, more accurate results
Automated profile generation for 3,600+ students with privacy-first design.
Tech Stack
Backend: FastAPI, PyTorch with CLIP, OpenAI API, Yale CAS (OAuth authentication)
Frontend: Next.js 14, TypeScript, Tailwind CSS, Framer Motion
Infrastructure: Railway (Docker), Vercel — sub-5s search response times
Key Features
- Natural Language Search — Find people using descriptive queries
- Advanced Filters — Filter by residential college, graduation year, and major
- Find Similar — Click any result to discover visually similar people
- Gamified Engagement — Leaderboards for trending searches and most-viewed profiles
- Anonymous Mode — Search privately without logging to history
- Content Moderation — AI-powered filtering using GPT-4o-mini