Operationalizing Vector Databases on Postgres

Come join us for an evening of Postgres and AI at the Computer History Museum with food and beverages sponsored by Tembo!

Abstract

Developers are increasingly generating embeddings from their operational data using Postgres, facilitated by tools like Pgvector which enable searching these embeddings. However, the challenge of building AI applications extends beyond storing and searching embeddings; there is a significant effort involved in generating and maintaining these embeddings to create a fully functional AI application.

In this talk, we will discuss the comprehensive workflow required to operationalize AI applications on Postgres. We'll explore the process beyond storing and searching embeddings, discussing advanced topics such as retrieval augmented generation and vector search. Additionally, we will address the nuances of updating embeddings as the source data evolves and lessons learned in building machine learning feature stores, highlighting how these lessons apply to developing AI applications.