GenAI for Business on the Modern Data Stack
summary
Argues that a Lakehouse plus GenAI becomes a data intelligence platform where Databricks layers Unity Catalog for governance, Delta/Parquet for scale and MosaicML (acquired for $1.3B) plus dbrx for LLMs. Describes four LLM phases: prompting → RAG → fine-tuning → pretraining, gives a demo where a code assistant auto-generates regex/SQL to clean movie titles and extracts gross values, cites training a custom model for ~$10–20K, and stresses lineage, LLMops, and serverless Databricks SQL for production use.
AI-written summary · about
chapters (7)
- 0:00Introduction and speaker background
- 7:14Data bricks and Lakehouse concept
- 14:42Modern data stack evolution
- 23:58Demo of GenAI in development
- 39:23Building custom LLMs
- 46:45Democratizing AI development
- 48:51Q&A and closing remarks
related talks
49mBuild & deploy generative AI apps in days with Databricks Lakehouse AI
24mMatei Zaharia, Databricks | Databricks Data + AI Summit 2023
26mJoel Minnick, Databricks | Databricks Data + AI Summit 2024
52mElevate your data and AI governance with Databricks Data Intelligence Platform
23mBuilding AI-Driven Real-Time Lakehouse APIs
14mTony Baer, dbinsight & Doug Henschen, Constellation Research | Databricks Data + AI Summit 2023
35mHigh-performance data foundation for AI success
07mBuilding and Deploying GenAI Apps at Block