Databricks as the Foundation for GenAI: What’s New and What Works
summary
Latest Databricks features and concrete behaviors are detailed: Agent Bricks (declarative agent patterns, built-in evaluation, beta and region-limited); Serverless GPU Compute for autoscaling training and serving; Vector Search that builds and serves vector indexes from Delta tables and managed or custom embeddings; Lakebase as a managed serverless Postgres for low‑latency OLTP; MLflow additions like the MLflow prompt registry and tracing; inference tables that log every request as Unity Catalog Delta tables and an AI Gateway that enforces guardrails.
AI-written summary · about
chapters (13)
- 0:00Databricks for GenAI intro
- 1:54Talk structure
- 2:59GenAI impact domains
- 4:26Information management shift
- 6:58Semantic management & vector search
- 12:06Model lifecycle evolution
- 15:39Process automation evolution
- 22:02GenAI evaluation challenges
- 24:33Governance for real-time AI
- 27:05Infrastructure for GenAI workloads
- 30:42Two-dimensional takeaways
- 33:11Summary
- 33:33Q&A
related talks
14mTony Baer, dbinsight & Doug Henschen, Constellation Research | Databricks Data + AI Summit 2023
24mMatei Zaharia, Databricks | Databricks Data + AI Summit 2023
02h 52mKeynote Day 1 - Full
03mDay 2 Opening Remarks
14mRik Tamm-Daniels, Informatica | Databricks Data + AI Summit 2024
20mSharad Kumar, Qlik | Databricks Data + AI Summit 2024
01mtheCUBE's Coverage of the Databricks Data + AI Summit
01h 21mKeynote on the "AI for Data" Revolution
