Srinivas Yadam

AI-Native Network Validation for Next-Generation Data Center Switches

Abstract:
Modern data center fabrics have evolved into AI-driven computational infrastructures operating at terabit speeds with programmable data planes, rendering traditional static validation methodologies fundamentally inadequate. Script-based regression suites built on fixed topologies and predictable traffic patterns cannot keep pace with environments defined by continuous reconfiguration, adaptive routing, and workload-driven topology mutations.

This talk introduces AI-Native Network Validation: a paradigm shift from passive test execution to autonomous cognitive systems that continuously learn, adapt, and self-optimize. The architecture combines reinforcement learning-based test orchestration, multi-layer telemetry ingestion, and high-fidelity digital twin environments to achieve validation coverage that static approaches cannot reach. Graph neural networks encode complex topological dependencies to guide intelligent scenario selection, while causal graph analytics enable automated root cause isolation across distributed infrastructure.

Attendees will explore how these systems address three critical deployment challenges: high-density AI fabrics with stringent tail-latency requirements, geographically distributed edge inference clusters demanding continuous rather than discrete validation, and disaggregated composable architectures introducing combinatorial configuration complexity beyond what human-designed test coverage can exhaustively address.

The session concludes with an assessment of open challenges including telemetry interface standardization, integration with legacy validation infrastructure, and federated learning frameworks for cross-organizational knowledge sharing without compromising proprietary data, and the path forward toward fully autonomous network assurance.

Profile:
Srinivas Yadam is a senior technical leader with 18+ years at Cisco Systems, specializing in cloud networking and data center technologies. As a leader in the System Integration Test Team for Cisco Nexus 9K DCBU, he drives critical product initiatives, ensuring high-quality deliveries that have generated significant revenue and customer acquisition.

His contributions include launching the Cisco Nexus 9364E-SG2-O 800G switch for AI/ML workloads, the industry's first 800G solution, generating $800 million in revenue. He has developed automation frameworks achieving over 70% test coverage and holds a USPTO-filed patent in context-aware AI-powered network diagnostics using NLP and LLM integration. Srinivas's expertise spans NX-OS, IOS-XR, BGP, MPLS, IPSec, NetFlow, streaming telemetry, and network programmability using Python, NETCONF, gRPC, and YANG modeling. His rigorous validation work has helped secure contracts with hyperscale operators including Microsoft, Google, Amazon, and Meta. He holds an M.Tech in Computer Science from IIT Guwahati and a B.Tech from Vignan's
Engineering College.