I build software by talking to it — then reading every line before it ships.

How I got here

I didn't come from a traditional engineering background where I spent years learning frameworks one at a time. What pulled me in was the opposite: watching the gap between having an idea and having a working product shrink from months to days. The first time I described a feature to Claude Code and watched it turn into a real, working screen, it stopped feeling like a novelty and started feeling like a genuinely different way to build software.

That's what "vibe coding" means to me — not typing less for its own sake, but spending almost all of my attention on the decisions that actually matter: what the product should do, how it should feel, and where it needs to be exactly right.

How it actually works

What I actually build

I build real, shipped products — not demos. My first is Invomly, an invoicing SaaS built specifically for French freelancers — with the authentication, billing, multi-language support, and compliance rules that a toy project would skip. I'm drawn to narrow, well-defined problems where I understand the user well enough to make good calls fast, and where AI-assisted development compresses the boring parts without compromising the parts that need real judgment.

What makes this different

"Vibe coding" has a reputation for meaning "ship whatever the model outputs and hope for the best." That's not how I work. Every diff gets read. Every edge case gets poked at manually before it goes live. The speed comes from compressing the loop between idea and working code — not from skipping the review that makes software trustworthy. I treat the AI as a fast, capable collaborator, not a replacement for judgment.