LLMs on Autopilot: Running AI Agents on Kubernetes With Open Source Tools
Talkat Open Source Summit Japan + AI_dev: Open Source GenAI & ML Summit + Automotive Linux Summit 2025 · 2025-12-08..12-10·
summary
Shows a practical stack that runs autonomous agents on Kubernetes using Crossplane for declarative infrastructure, Prometheus/cloudwatch for metrics, and Claude for reasoning, with Ray suggested for heavy distributed workloads. Demonstrated a recorded control loop that reads RDS metrics every 10 minutes from a Crossplane resource, prompts Claude to decide, and updates desired state causing an automatic scale from T3 micro to T3 medium. Includes policy guardrails, Kyverno and Falco for safety.
AI-written summary · about
chapters (18)
- 0:00AI agents on Kubernetes
- 3:17What are AI agents
- 7:50AI agents vs humans
- 12:48Why platforms fail agents
- 15:59Essential agent ops elements
- 19:08Future-proof tools
- 20:35Crossplane for interaction
- 23:26Prometheus perception layer
- 24:30Demo: autonomous agent
- 29:44Demo results workflow
- 30:06CNCF AI landscape
- 31:29Layered agent tooling
- 32:11SRE and security agents
- 33:15AI agent deployment tips
- 35:23Preventive and detective guardrails
- 37:50AI agent infrastructure needs
- 38:32Resources and reading
- 39:58Q&A security prompts
related talks
30mLLMs 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
34mDevOps at LLM Speed: Using an AI Copilot for Kubernetes and Jenkins
30mSo You Want To Run AI Agents on Kubernetes: A 101 Guide
32mBuild your own Autonomous Agent powered by LLMs using Kubernetes
14mReasoning Operators: Bringing LLM Logic Into Kubernetes Control Loops
40mAgents in Action: Advancing Open Source AI Missions