Programmer at heart, architect by conviction.

I learned to code in my early teens, in the era when the web was still held together with view-source and curiosity. Twenty-some years later I'm a tech lead at HalfSerious in Montréal.

My work sits at the intersection engineering organizations are all being pulled toward: how humans and AI build software together. I author and own our human-AI engineering methodology — the development workflow, the prompt library and reviewer personas, and the threat modelling that has to hold underneath it. Alongside it I design the CI enforcement that makes architectural invariants mechanical rather than vigilance-dependent. The thesis running through everything: humans as architects of intent, AI as executor. The first half of that phrase is Bornet et al.'s, from The Human-Agent Orchestrator (opens in a new tab) — I've been building the practice underneath it since. Most of what I work on starts that way: read the research, argue with it, and find out what it costs to make it operational.

Before this, I led delivery teams for enterprise clients — building a scrum team from scratch, running one-on-ones and evaluation cycles. I'm also tech lead on NIC, an AI-agent platform built for ArcelorMittal Canada's mining operations, which I took into production and still maintain. Its agents recommend procedures to the humans monitoring the fleet — and decide nothing on their own. Les Affaires reported (opens in a new tab) on one of them catching an oil leak on a haul truck before it became an engine failure and an environmental spill (in French). That combination — leading delivery teams before, staff-track IC now — is deliberate: I chose this path with full knowledge of both.

The rest of the time: systems thinking, organizational design, the tension between process and judgment, AI capability forecasting — and two cockatiels who supervise all of it.