The journey
A conviction, since childhood.
I · Origin
I was eight when it began — though I had no words for it yet. My father, a CFO, would go to the office on weekends to deploy AS/400 systems, and I tagged along. I watched the terminals, the quiet discipline of machines that simply had to work because people and businesses depended on them. What I took from him wasn’t only a fascination with technology — it was resourcefulness: figuring things out, repairing, adapting, and never standing still, even when the path wasn’t obvious.
Very early, I felt computers were powerful but built the wrong way around. We had to type, click, search — to bend ourselves to the machine. I wanted the opposite: machines that understand us, starting with the most human interface we have — language. As a child, I was captivated by Ulysses 31, where the ship’s computer spoke with its crew. That talking machine planted a seed that never left me — and set me on the path to dialogue systems. The principle has never changed since: hide the complexity behind the simplest gesture there is — talking to the machine.
II · Conviction
For close to seventeen years, I carried that vision almost alone. People around me said it wouldn’t work — the technology wasn’t ready, the market wasn’t ready, the world wasn’t interested. I wasn’t waiting for AI to become fashionable. I kept building toward something I was certain had to exist.
III · Proof
2012. Microsoft Days, Paris — the third day, dedicated to innovation and research. Three thousand people in the room, twelve thousand more online. I was there to present the work of our research lab in Lille, xBrainSoft: Angie, the assistant the press would soon call the “French Siri.” The demo was, in effect, a ChatGPT experience ten years early — a pure spoken dialogue between a human and a machine: conversation, tasks, actions. Introducing me on stage: Bernard Ourghanlian, CTO of Microsoft France and my mentor — one of the first to believe in my work on AI.
As Bernard walked on to introduce me, I ran one last check. The Wi-Fi was gone. Three thousand people had tried to connect at the same time, and nobody had planned for it. Every part of the demo ran live, over the network. Backstage, panic — I still apologize to the crew for those few minutes. I honestly thought my career in AI was ending right there. And Bernard was already calling me up. Double or nothing: either it worked, or it was over.
I walked out and asked the whole room to switch off their Wi-Fi — “imagine you’re inside a giant 747.” Three thousand people did. The connection came back. And Angie came alive: translating on command, finding the cheapest fuel nearby, and finally — unprompted — reminding me of a meeting at Bastille in thirty-five minutes. Everything worked. The room applauded.
That moment changed my life, fast. Three months later I was in Silicon Valley, opening my own research lab in dialogue systems and NLP. The rest is history.
Watch the moment ↗
IV · Direction
Since then I’ve had the privilege to work with governments, global companies, universities and scientific institutions — including NASA. But the deeper lesson came later. As AI grew more powerful, I saw the risk: built in the wrong direction, it can replace more than it augments, centralize intelligence instead of sharing it, make people more dependent instead of more capable.
“AI won’t replace you. A person using AI will.”
2016 — six years before ChatGPT
So my work today isn’t only about building AI systems — it’s about building them in the right direction: to help people think, learn, create and decide with more dignity. Not just experts. Not just executives. Not one country or one industry. Everyone. That is why so much of what I build is open source, why I teach, and why I commit to non-profit work: democratizing has to happen in how AI is built and shared — not just in how we talk about it.
“If AI is rewriting the rules of productivity, we must rewrite our relationship with work itself.”
2024 — Work in the Age of AI, Part 1
V · Craft
Through all of it, I never left the keyboard. I could have become a manager and moved into pure strategy; instead, I chose to keep building. I’m convinced you only truly understand AI by making it work under real constraints, with your own hands — and that’s exactly where my method comes from: not theory, but thousands of hours shipping systems that had to run in the real world. I’m an architect who still builds.
The best technology isn’t the one that replaces us. It’s the one that helps more people become capable.
