LLMs on Autopilot: Running AI Agents on Kubernetes with Open Source Tools
summary
Shows a working control-plane pattern that runs open-source LLM agents on Kubernetes to observe, reason, and act: metrics from Prometheus/CloudWatch feed an LLM (Claude) that decides to change a Crossplane resource and Crossplane reconciles it. Demo auto-scales an AWS RDS MariaDB from t3.micro to t3.medium via a Crossplane cron operation every 10 minutes; scaling completes successfully. Uses Kyverno/Falco for guardrails and Ray for heavier distributed reasoning.
AI-written summary · about
chapters (12)
- 0:00Introduction and speaker intro
- 1:45AI impact on infrastructure
- 3:29What are AI agents
- 5:54Agent requirements for infrastructure
- 10:50Agentic ops requirements
- 12:14Crossplane for AI agents
- 14:00Stages of AI-enhanced control
- 15:22Demo autonomous scaling agent
- 18:49AI agent tooling landscape
- 21:16Tips for building agents
- 23:45Summary and resources
- 25:30Q&A session
related talks
42mLLMs on Autopilot: Running AI Agents on Kubernetes With Open Source Tools
10mLightning Talk: LLM-Driven Applications: Intro to LLM Agents in Kubernetes
25mLLM Powered Agents with Kubernetes
08mKeynote: Plug in and Scale: Serving LLM Models on Kubernetes Ma...
31mCluster Operations as a Service, Introducing LLM Backed Controllers.
40mAgents in Action: Advancing Open Source AI Missions
14mReasoning Operators: Bringing LLM Logic Into Kubernetes Control Loops
50mRunning Open LLM Models on Kubernetes