The Everything Feed - All Packet Pushers Pods HN800: Root Cause Analysis for the Entire Stack (Sponsored)
Oct 10, 2025
Join John Capobianco, Head of Developer Relations at Selector, and Vreja Sriram, Head of Application and Data Engineering, as they dive into the intriguing world of AI-driven full-stack root cause analysis. They discuss how Selector customizes its models for clients, linking diverse systems to identify issues efficiently. Learn about a real-life case of Kubernetes auto-remediation and how they dramatically reduced support tickets by correlating data. Plus, discover the power of natural language queries for seamless insights!
AI Snips
Chapters
Transcript
Episode notes
Models Need Your Context
- Selector builds a data-centric model tied to each customer's unique metadata and relationships.
- Correlated, actionable insights depend more on meta tags than raw device metrics.
Tag Your Infrastructure Properly
- Define and supply human-friendly meta tags like region, city, and building to improve relevance.
- Use those tags to enable accurate natural-language queries and crisis summaries.
Ensemble Models Power Outcomes
- Selector is an ensemble of models: LLMs for language, regression for baselining, clustering for patterns, and temporal correlation.
- The platform's models stay constant but continually learn from customer data.