Dinesh Thangaraju
Head of AWS Data Platform, Amazon Web Services

As organizations race to deploy AI agents across their enterprises, they’re encountering a critical bottleneck: fragmented knowledge bases and siloed agent capabilities that prevent comprehensive, cross-domain insights. This presentation explores how to address this challenge through a federated approach to enterprise agentic AI infrastructure. This session examines three fundamental challenges facing enterprise AI adoption:

The Knowledge Fragmentation Problem: Organizations are independently building knowledge bases for specific use cases, leading to duplicated effort, inconsistent data parsing strategies, and context drift as source documents evolve. Each team creates their own chunking, embedding, and retrieval mechanisms without a unified architecture—resulting in AI agents that cannot access cross-functional domain knowledge or maintain accuracy at scale.

The Agent Collaboration Gap: Today’s AI agents operate in isolation within their domains. When business leaders ask complex questions spanning customers, pricing, services, and operations, they receive fragmented answers requiring manual synthesis. Without standardized agent-to-agent communication mechanisms, authentication frameworks, or centralized agent registries, organizations cannot deliver the holistic insights that drive strategic decision-making.

The Governance and Innovation Paradox: Enterprises need to enable rapid experimentation while maintaining security, compliance, and user experience standards. Traditional centralized approaches stifle innovation; purely decentralized approaches create chaos. The challenge is building federated frameworks that guide discovery, assessment, incubation, and graduation of AI solutions without creating bottlenecks.

This presentation introduces a three-pillar architecture for enterprise agentic AI:

Unified Federated Knowledge Base: A bottom-up approach where domain teams create specialized knowledge bases that integrate into an organization-wide ecosystem through standardized ingestion pipelines, interfaces for vector databases and knowledge graphs, and evaluation frameworks for accuracy and relevance.

Cross-Domain Agent Collaboration: Technical mechanisms enabling AI agents to discover, authenticate, and communicate with each other—transforming isolated data points into comprehensive business intelligence that spans organizational boundaries.

Federated Innovation Framework: Secure sandbox environments with structured governance models that accelerate development time by 50% while maintaining enterprise standards for security and user experience.

Attendees will learn:

• Practical patterns for building federated knowledge architectures that eliminate redundant development efforts
• Technical approaches to agent-to-agent communication, including authentication, authorization, and service discovery
• Governance models that balance innovation velocity with enterprise security and compliance requirements
• Implementation strategies for co-owned initiatives requiring commitment across multiple data organizations

This session is essential for Chief Data Officers, enterprise architects, and data leaders navigating the transition from siloed AI experiments to scalable, federated agentic AI ecosystems that deliver measurable business value.

Back To Proseminar Series