Pi Multi-Agent Teams - Beyond the YAML Basics
Building multi-agent teams on Pi with custom extensions. Role design, communication topology, model-per-agent strategy, convergence budgets, and the TUI widget that shows where your money is going.
Engineering leader. Twenty-five years in software, twenty of them leading teams, currently Senior Director of Engineering, Platform at Entrust in identity verification.
This site is the workshop, not the CV. I build agent systems and small tools here, and write about what actually breaks when you do.
More about me and how I got here

Coded, multi-step agent workflows against real repositories, with a board that shows what is happening while it is still happening. Under active development, and the interesting parts are the constraints.

Scan a supermarket barcode with your phone and get a 0-100 health score that shows its working. Portuguese-first, built on Open Food Facts, and early enough that the methodology is still the product.

This website. A Next.js 16 blog with MDX posts, full-text search, tag pages, dark/light themes, and Playwright end-to-end tests running in CI.
Building multi-agent teams on Pi with custom extensions. Role design, communication topology, model-per-agent strategy, convergence budgets, and the TUI widget that shows where your money is going.
Extensions change what Pi can do. The TUI changes what Pi feels like to use. Build a custom footer with live status, a personal theme, and a prompt-picker overlay - and learn where the customization surface ends.
Skills teach the LLM what to do. Extensions teach Pi what to become. Build a safety gate, a custom slash command, and a context injector - and learn when to reach for code over markdown.