Oleksandra Bovkun | Data Track | Compass AI & Data Summit 2025
summary
Databricks announces Lakebase as a Postgres-compatible OLTP layer for the Lakehouse that separates storage and compute, uses object storage (S3/ADLS), and exposes branching for cheap schema/dev branches. The talk demos a chatbot that stores conversation memory in Lakebase, shows an automated pipeline syncing the OLTP tables into Unity Catalog for analytics, and verifies queries over the synced data — the demo works end to end. Practical claims: fully managed Postgres API, reverse-ETL and model-serving patterns supported, and cost-efficient scaling for many small AI apps.
AI-written summary · about
chapters (10)
- 0:00OLTP in the lakehouse
- 1:33OLTP for AI applications
- 8:19OLTP vs OLAP in AI
- 10:29Combining operational and analytical data
- 14:01Lakehouse for AI applications
- 18:20Lakebase architecture and DX
- 19:26Live chatbot demo with memory
- 28:43Lakebase use cases in AI
- 30:29Unifying operational and analytical data
- 32:35Q&A
related talks
28mLakehouse Architecture with Delta Lake and Databricks
44mWhat's the Best Big Data Architecture for You?
34mAnalytics and AI with PostgreSQL - Mission Possible!
25mBuild open format data lakehouse
23mBuilding AI-Driven Real-Time Lakehouse APIs
52mElevate your data and AI governance with Databricks Data Intelligence Platform
42mBuilding a Lakehouse data platform using Delta Lake, PySpark, and Trino
29mLakeside AI: Bringing Models to Data with Open Lakehouses & Agents)