DevOps engineer. I've spent my career building enterprise software and delivery pipelines, with testing as my craft. Now I put AI agents to work on all of it.
- frank: a self-hosted Kubernetes cluster on heterogeneous hardware, immutable OS, fully declarative. My first large-scale AI-native project.
- blog.derio.net: the cluster's blog, written in the voice of Frank himself: his build, his operations, and the decisions that shaped him.
- super-fr: a Claude Code plugin suite for autonomous feature delivery: brainstorm to reviewed PR inside isolated workspaces (git worktree + devcontainer), with phase dispatch to background runners.
- vscode-launchpad: a Tauri dashboard to search and open any VS Code workspace, with a live monitor for Claude Code sessions and a kill switch for the zombies. Built spec-first with openspec.
- blog-craft: portable scaffolding and authoring skills for teaching blogs, built as Claude Code skills.
Multi-agent software delivery under real engineering discipline: spec-first, tight TDD, cheap-to-iterate architecture. What I'm after is the layer underneath: agent swarms that are trustworthy (reliable, secure, autonomous at scale) and composable (workspaces, skills, and secrets tunable per agent). I haven't found a system that does both well, so I'm building my own.
Kubernetes · Talos · ArgoCD · GitOps · Terraform · Python · TypeScript · platform engineering · TDD/BDD and test automation · multi-agent systems