The power of AI and Copilot for Azure Databases
summary
Announces vector search in Azure Cosmos DB using DiskANN (compressed vectors in memory, full-fidelity graph on SSD) with stable recall under heavy inserts/deletes and enterprise-scale usage already in Bing; PostgreSQL adds local E5 embeddings inside the engine (no external calls), demo shows a 384-dimension local query at ~4 ms and a stress run hitting 242 TPS versus ~15–20 TPS for remote embeddings; PGVector compatibility, DiskANN coming to PostgreSQL before summer, and database Copilots that suggest indexes, surface blocking chains, and convert natural language to SQL.
AI-written summary · about
chapters (10)
- 0:00Welcome
- 1:05AI in modern databases
- 9:22Vector search and RAG patterns
- 17:28DiskANN in Cosmos DB
- 22:46Demo: Local embeddings in PostgreSQL
- 27:03PostgreSQL index recommendations
- 28:05Developer experience with AI
- 31:19Azure SQL Copilot demo
- 40:21Copilot for SQL Server & MySQL
- 45:39Closing & future of AI
related talks
15mAzure Cosmos DB as Vector Database
27m6 things you can do with azure_ai & PostgreSQL on Azure
50mHow to design and build AI applications with vector search using Azure OpenAI & Azure Cosmos DB
19mHow Microsoft powers planet-scale AI apps with DiskANN
40mUnleashing the Power of Vector Search with Azure SQL Database
03h 08mAzure Cosmos DB Conf 2025
12mBuilding Generative AI Apps with Azure Cosmos DB
03h 12mAzure Cosmos DB Conf 2024