It's built using Neon Postgres + pg_embedding and OpenAI for generating embeddings. More details can be found in the repo https://github.com/neondatabase/yc-idea-matcher