Evaluating and Tracing a Multi Modal RAG Application

As AI systems incorporate multiple modalities like text, images, and audio, multi-modal Retrieval-Augmented Generation (RAG) applications are becoming increasingly prevalent. However, building and deploying these complex models presents unique challenges. In this session, Hakan Tekgul--Solutions Architect at Arize AI--explores the key distinctions between multi-modal RAG applications and traditional RAG models, highlighting best practices for their development and deployment. By leveraging Arize Phoenix, an open-source tracing and evaluation tool, Hakan dives into evaluation techniques tailored to multi-modal systems, and examine essential troubleshooting tools and workflows to diagnose and resolve issues that may arise. Join this session to unlock the full potential of multi-modal RAG models. This talk was originally delivered at Arize:Observe 2024 at Shack 15 in San Francisco on July 11, 2024.