Mr. Aswathnarayan Muthukrishnan Kirubakaran

Distributed Retrieval Augmented Generation (RAG) for Large Scale Enterprise Intelligence Systems

Abstract:

Retrieval Augmented Generation has become a key approach for enterprise intelligence by allowing large language models to reason over proprietary data while maintaining governance and security. However, traditional centralized RAG architectures face core challenges when used at enterprise scale, including cross domain access bottlenecks, storage duplication, compliance risks, and rising latency across heterogeneous data platforms.

This invited talk presents a distributed RAG architecture designed for real enterprise environments that span multi cloud infrastructures and fragmented knowledge repositories. The system separates document ownership from retrieval execution, enabling localized vector indexing, metadata driven routing, and federated ranking across independent enterprise data zones. Experimental evaluations show strong retrieval accuracy that is comparable to centralized RAG while reducing data movement cost, improving tenant isolation, and lowering inference latency. The talk will outline the architectural foundations, describe deployment considerations, and highlight future directions for scalable and trustworthy enterprise RAG systems.

Profile:

Aswathnarayan Muthukrishnan Kirubakaran is a Senior Data Engineer and AI researcher specializing in large scale data platforms, distributed intelligence systems, and sensor analytics. He works at Meta Reality Labs in California, developing high performance telemetry pipelines and data models for wearable and neural interface devices using EMG signals. His prior work includes deep learning research in medical imaging at Agfa Healthcare, the design of mammography screening prioritization models, and data automation frameworks. He is an IEEE Senior Member and contributes to research, technical reviewing, and industry collaborations.