Creating a Retrieval Augmented Generation (RAG) system with a few pdf documents is easy. But, building a RAG system using enterprise data sources like document hubs, ticketing systems and collaboration tools is challenging. Each enterprise has its own vocabulary that is not modeled by standard embedding models. Out-of-the-box RAG techniques do not fully address the temporal requirements for enterprises. In addition, strict access controls and authorization schemes implemented in the data sources are hard to replicate when creating multiuser RAG systems with those sources. How do we solve RAG in each of these dimensions? 

In this session, Kumaran will discuss techniques that help overcome these challenges in enterprise RAG with multi-dimensional retrieval. He will walkthrough the challenges faced, experiments done, solutions implemented and improvements obtained with multi-dimensional retrieval. This session will help managers and engineers working on building RAG systems for their enterprise use cases.