Joe Sack
Product leader for query engines and AI search
I led product for the SQL Server query optimizer at Microsoft, where the features I shipped now run in billions of Azure SQL queries a day. I went on to lead the query engine and Atlas Search teams at MongoDB, Azure SQL Copilot and natural-language-to-SQL at Microsoft, and the Elasticsearch core platform at Elastic. I still build hands-on.
Billions of queries a day
SQL Server query optimizer features I shipped at Microsoft, including Adaptive Joins, Memory Grant Feedback, and Parameter Sensitive Plan Optimization.
Search and vector at scale
Led the MongoDB query engine and Atlas Search, and the Elasticsearch core platform, including vector and hybrid retrieval.
20+ years in query engines
From the SQL Server optimizer to AI search.
Vector Search Essentials
New CourseA comprehensive course preparing SQL Server DBAs and Developers for AI-powered semantic search. Covers vector embeddings, DiskANN indexing, hybrid search, and production deployment.
View course details →Vector Search in Practice
New CourseProduction patterns for real-world vector search, building on Essentials. Covers embedding selection, search quality metrics, reranking, RAG, and multi-modal search.
View course details →Earlier training: 13 SQL Server courses on Pluralsight →
Earlier books: four SQL Server titles from Apress →
Earlier paper: the SQL Server 2014 Cardinality Estimator whitepaper →
Selected writing
I write about agentic-first databases, where AI agents, not people, write the schemas and the queries.
Agentic-First Databases in 2028
What databases look like when AI agents become the primary users.
The Dimension Reduction Cliff
Smaller vectors, different results. How much drift can your use case tolerate?
From 15th Place to Gold Medal
How semantic reranking transforms vector search results.
Four Numbers That Tell You If Vector Search Works
Precision, recall, MRR, and nDCG: the metrics that tell you if vector search works.
Patent
US Patent 11,334,538
System and method for cardinality estimation feedback loops in query processing. Co-invented at Microsoft.
View patent →