The Story So Far
I spent most of my career inside enterprise engineering at BNP Paribas Fortis. I worked as a senior developer and later as Technical Lead for the Enablement project within Centric, a financial platform serving more than 100,000 users.
Alongside that work, I built DinnerGift from scratch and ran it as a commercial platform — owning the backend, frontend, architecture, deployment, and long-term maintenance.
Those environments taught me what production responsibility actually means: making trade-offs under delivery pressure, keeping systems understandable over time, and knowing when a technically plausible solution is not yet safe to ship.
The Shift
Over the last two years, I have been exploring what changes when AI becomes part of the everyday engineering workflow — not as an autocomplete tool, but as a collaborator that can research, implement, test, review, and document.
I built a workflow in which AI agents handle much of that volume while I remain responsible for architecture, security, quality, and release decisions. The most important lessons have not come from the tasks the agents complete successfully, but from learning where they sound confident while being wrong.
That workflow now shapes how I work. When I join a team as a developer or tech lead, I can also help introduce the useful parts of it — adapted to the team's codebase, constraints, and existing practices.
The tools behind my workflow
I run Hermes Agent, an open-source autonomous AI coding platform, as my core development tool. It operates in a persistent loop: receiving tasks, reasoning through them, writing and executing code, reading test output, and self-correcting. The open-weight model is served through Ollama Cloud. Ollama states that prompt and response content is not stored or logged and is not used for training; vendor documentation should be reviewed against the requirements of each engagement.
Alongside Hermes, I use Claude for visual design review — it can see and score a screenshot, but never touches code. Two AIs, two roles: one implements, one critiques.
What This Means for You
I begin with the actual work: understanding the codebase, delivery pressure, technical constraints, and the way the team collaborates.
From that position, I can help improve both the software and the workflow around it. That may include architectural guidance, hands-on implementation, stronger review practices, or introducing AI support where it removes repetitive work without weakening accountability.
The goal is not to maximize AI usage. The goal is to help the team deliver good software more consistently.
If that sounds like what you need, see how I work.