In episode 119 of the Acceleration Economy Minute, I share an update on Databricks and a new toolkit for driving large language model (LLM) apps.
Highlights
00:18 — Databricks has announced a new Retrieval Augmented Generation (RAG) toolkit
for its data intelligence platform that will help customers build, launch, and run large language model (LLM) driven apps for diverse use cases.
01:01 — The new Databricks tools make it easier for companies to create RAG apps which are notoriously difficult to develop because they require collecting data — structured or unstructured — from multiple sources. The data then needs to go through a preparation process to make it actionable. The tools currently in public preview include:
- A vector search service for semantic search on existing tables
- Fully managed foundation models providing paper token-based LLMs
- A quality monitoring interface for users to oversee production performance
- Development tools to compare and evaluate various LLMs
