Session

Talking to All Your Data: Building Multi-Agent Systems for Structured and Unstructured Information

Overview

ExperienceIn Person
TypeBreakout
TrackArtificial Intelligence
IndustryEnterprise Technology
TechnologiesMLFlow, Mosaic AI
Skill LevelIntermediate
Duration40 min

Learn how to build sophisticated systems that enable natural language interactions with both your structured databases and unstructured document collections. This session explores advanced techniques for creating unified and governed AI systems that can seamlessly interpret questions, retrieve relevant information and generate accurate answers across your entire data ecosystem.

 

Key takeaways include:

  • Strategies for combining vector search over unstructured documents with retrieval from structured databases
  • Techniques for optimizing unstructured data processing through effective parsing, metadata enrichment and intelligent chunking
  • Methods for integrating different retrieval mechanisms while ensuring consistent data governance and security
  • Practical approaches for evaluating and improving KBQA system quality through automated and human feedback

Session Speakers

IMAGE COMING SOON

Xiangrui Meng

/Principal Software Engineer
Databricks

IMAGE COMING SOON

Kasey Uhlenhuth

/Director of Product
Databricks