I help engineering teams put AI to work without making their codebases worse.

I'm Patrick Ryll. For twelve years I've architected and delivered enterprise web applications across finance, logistics, healthcare, education, and government — most recently as a senior front-end architect specializing in React, TypeScript, and Next.js, with full-stack work in Node.js, APIs, cloud services, and databases.

What occupies me now is what happens when an engineering team meets AI coding tools. The pattern keeps repeating: AI refactoring is additive, so it extends whatever logic it finds. In a legacy codebase whose internal logic is already compromised, that produces more code, more surface area, and more debt — generated faster than anyone can review it.

So the work is redirecting duplicative and shared coding tasks to agentic tools, and moving human attention up: to architecture, to review, and to deeper conversation with the business groups the software actually serves. Not developer-by-developer heroics against an outdated codebase, but a deliberate division of labor.

Before software I trained as a lawyer — a Juris Doctor from New England School of Law and an LLM in Taxation from Northwestern. It turns out to be unusually useful. Reading a regulation closely, arguing from a specification, and being precise about what a system is obligated to do are the same discipline in both professions, and it showed in the work for the Department of Veterans Affairs, for Wells Fargo, and for federal assessment programs.

Permacodeture is a pun on permaculture, which I practice. Permaculture designs systems that keep producing without constant intervention: build the soil, observe before acting, place things where they support each other, and let succession do the work you would otherwise do by hand. That is a fair description of what I try to do with a codebase and with an engineering team.