How we built a SOTA search engine using PostgreSQL, pgvector, and Qwen3 embeddings [P]

I wrote a technical breakdown of how search works on Papers with Code. The system combines keyword and semantic search, which produced better results than either approach alone. The stack includes: PostgreSQL with pgvector Qwen3-Embedding-0.6B for text embeddings Hugging Face Jobs with an NVIDIA L4 for batch embedding generation Hugging Face Buckets for storing artifacts A live embedding model…

See it on reddit.com ↗