Santosh Kumar Vangapelli

Talk Title:

When AI Becomes Your Best SRE: Rethinking Root Cause Analysis in Real-Time Distributed Systems

Abstract:

Modern enterprises operate distributed systems spanning many interacting services, asynchronous event streams, and multi-layer infrastructure. Yet when something breaks, the visible symptom is seldom where the fault originated. A delayed order may trace back to a schema mismatch, a retry amplification loop, or a degraded dependency buried beneath layers of compensating behavior. The core problem isn't tooling. It's that operational truth is fragmented across time, ownership boundaries, and technical layers that no single dashboard, log, or trace can fully capture. This session confronts the hard reality: traditional root cause analysis built on alert triage, manual log correlation, and the intuition of experienced operators breaks down at scale. Secondary symptoms dominate the operational surface, alert inflation increases cognitive load without increasing causal clarity, and accumulated expert intuition cannot be scaled or automated. AI changes the equation, but only under specific conditions. Drawing from real-world distributed system architectures, this talk outlines exactly what needs to be true before AI-driven diagnosis becomes reliable rather than a liability. We'll cover five non-negotiable platform requirements: trustworthy service-state context, task-oriented diagnostic interfaces, machine-readable schema and change awareness, structured operational memory, and policy-governed access boundaries. Attendees will leave with a concrete architectural pattern for AI-assisted diagnosis, one that correlates live telemetry with historical incident patterns, surfaces causal hypotheses with supporting evidence, and delivers structured next steps without collapsing human oversight. Whether you're scaling an engineering org, building internal platforms, or evaluating AI investments in your SRE stack, this session offers a grounded, implementation-ready framework for making AI a genuine diagnostic accelerator and not just another source of operational noise.

Brief Profile:

Santosh Kumar Vangapelli is a Staff Software Engineer and platform leader with over 12 years of experience building and scaling distributed systems, ML-enabled platforms, and high-throughput infrastructure across some of the world's most demanding technical environments. Currently at Coupang as a Staff II Engineer, Santosh founded and leads the company's digital twin simulation platform, a system designed for large-scale fulfillment optimization, scenario planning, and high-stakes operational decisions. He also led the re-architecture of Coupang's mission-critical inventory platform, significantly improving scalability and latency and supporting systems that handle orders at massive scale across Coupang's fulfillment network. Before Coupang, Santosh was a Member of Technical Staff at eBay, where he designed and shipped embedding-based image search, launched ad recommendation APIs, and built eBayGPT, an internal LLM-powered developer assistant that improved engineering productivity and workflow efficiency. At Lyft's self-driving division, he developed real-time LiDAR perception pipelines and optimized data labeling workflows that delivered significant cost savings. Earlier in his career, Santosh co-founded MoBolt, a mobile recruiting SaaS platform he scaled from the ground up before it was acquired by Indeed. He also held an engineering role at Google, working on enterprise search and test infrastructure. Santosh holds a B.Tech in Computer Science from IIT Kanpur, a highly competitive national program, and an M.S. in Computer Science from UMass Amherst. He has been recognized with Coupang's Top Tier performance rating (top ~5%) in both 2024 and 2025.