AI-Powered Root Cause Analysis at Scale: From Theory To Production... Letícia Mota & Yevgeny Gladun

This technical deep-dive from KubeCon presents Nubank's production AI-powered SRE Agent that automates root cause analysis across their 120M+ customer infrastructure. The talk addresses the critical challenge of processing 23,000+ Prometheus metrics through an "Expert Guides" pattern that intelligently filters to ~14 relevant data points before LLM processing, using Model Context Protocol for pluggable multi-source observability orchestration. Ideal for SREs, platform engineers, and anyone building AI-driven incident response systems at scale.