Production support using AI

Production support using AI Discover how to transform your production support engineering using AI to enable faster root-cause analysis and reduce investigation times for engineers of all experience levels. In this session, speaker Abhishek Kumar breaks down the traditional challenges of investigating system failures across complex distributed architectures and demonstrates how AI agents can bridge the knowledge gap. Learn how to combine static workflows, live infrastructure context, and historical session learnings to empower your support teams. Key Topics Covered: The Investigation Challenge: Why traditional on-call support takes significant time for new engineers and how context builds over time. Core Context Pillars: Breaking down production support into Static Context, Live Data Context, and Learned Context. Securing Live Context with MCP: How to set up a Model Context Protocol (MCP) gateway inside your cluster to securely query live data (logs, traces, databases) with read-only permissions and full auditability. Optimizing Static Context: Storing system flow architectures in concise YAML files to minimize token consumption and maximize context efficiency for AI agents. Building Learned Context: Leveraging past Root Cause Analysis (RCA) session files, handling archive caps to protect context windows, and continuous backlog improvement for code and observability. About the Speaker Abhishek Kumar is a Solution Consultant at Sahaj, working on a large-scale AdTech platform spanning multiple regions. Recently, he's been using AI agents to investigate production issues across infrastructure, turning tribal knowledge into a system that learns from every investigation. #productionsupport #AIinDevOps #modelcontextprotocol #SystemArchitecture #SoftwareEngineering #DevOps #AIAgents