From Toil To Triumph: Harnessing Agentic AI To Streamline Infrastructure as Code
Talkat Cloud Native + Kubernetes AI Day Europe 2025 · on Tuesday, 2025-04-01· at 15:20· Level 1 | Hall Entrance N10 | Room G·
summary
Builds a GitHub-integrated reviewer for OpenTofu/Terraform pull requests that runs on a production copy of a repo and evaluates across 700+ pull requests, using a hierarchical supervisor with seven agents (PR title/description reviewer, static analyzer, code-review team, code consolidator) orchestrated by LangGraph and calling models plus deterministic tools. Results are mixed but promising: agents post automated PR comments, catch duplicate AWS resources, produce better docs, and are validated with humans and LLM scoring; reusable agent publishing aims for an internet of agents.
AI-written summary · about
chapters (12)
- 0:00Agentic AI for infrastructure
- 2:29Internet of Agents
- 7:03Infrastructure as Code problems
- 10:14Agentic applications
- 14:11Task-specific agents
- 16:50Agentic review system
- 19:39Live demo: Agentic reviewer
- 25:44Other agentic workflows
- 27:19Evaluating agent performance
- 28:45Key learnings and future steps
- 30:50Future of agent interoperability
- 31:31Q&A
related talks
37mAI-Driven Developer Productivity at Scale
09mImproving code reviews with AI agents
30mBuilding Tools with Tools: Modern AI-Driven Development
46mAgentic Engineering: From Writing Code to Reviewing It
36mAgentic CI/CD: from Pull Request to Production without the Paper Cuts
40mHow To Tame Your Agents?
25mFrom AI Chaos to Production Ready: Using AI Agents to Improve Your Code Quality
24mEmpowering the GitOps Future with AI