Databricks expands Mosaic AI to help enterprises build with LLMs

The Mosaic AI Agent Framework utilizes Databricks' serverless vector search functionality by providing developers with the tools to build their own Retrieval Augmented Generation (RAG)-based applications on top of it34. The framework takes advantage of the hybrid approach of Databricks' vector search system, which combines classic keyword-based search with embedding search. This integration allows for the creation of compound AI systems that offer higher production quality and accuracy. The serverless vector search functionality is deeply integrated with the Databricks data lake, ensuring that data on both platforms is always automatically kept in sync and governed by the Databricks Unity Catalog governance layer14.

Databricks is launching five new Mosaic AI tools at its Data + AI Summit1. These are:

1. Mosaic AI Agent Framework: This framework allows developers to build their own retrieval augmented generation (RAG)-based applications on top of Databricks' serverless vector search functionality.

2. Mosaic AI Agent Evaluation: This is an AI-assisted evaluation tool that uses LLM-based judges to test the performance of AI applications in production. It also allows enterprises to quickly get feedback from users.

3. Mosaic AI Tools Catalog: This catalog makes AI services more discoverable across an organization and allows enterprises to govern which AI tools and functions their LLMs can call upon when generating answers1.

4. Mosaic AI Model Training: This service allows users to fine-tune models with their organization's private data to help them perform better on specific tasks.

5. Mosaic AI Gateway: This is a unified interface to query, manage, and deploy any open source or proprietary model. It allows users to query any LLM in a governed way, using a centralized credentials store.