Postgres and OpenAI - a match made in Azure - Giorgi Dalakishvili - NDC London 2026

A conference talk demonstrating how to run AI workloads directly from PostgreSQL on Azure using the Azure AI extension. The talk covers four use cases: semantic search using vector embeddings and pgvector, document summarization (both extractive and abstractive), sentiment analysis, and PII detection. All demos call Azure OpenAI and Azure AI Language Services directly from SQL queries, eliminating the need for separate application-layer glue code. Key concepts include vector embeddings, cosine similarity, KNN vs approximate nearest neighbor search, and the DiskANN indexing strategy for scalable vector search.