What I do
I run independent experiments on autonomous AI systems.
Companies are spending millions on AI agents without knowing whether they'll work for their specific use case. I build working prototypes, stress-test them, document everything - including the failures - and deliver honest answers.
My approach:
- Experiment over theory - I build the system, run it, measure what happens
- Failures are findings - every broken thing is a lesson, not a liability
- Research-grade documentation - you get reports, not slides
- No vendor alignment - I work for you, not for an AI platform
What I offer:
- Sponsored Experiments - You bring the question, I build the system and deliver a research report.
- AI Agent Readiness Assessment - Two-week audit: which of your processes are viable for agent automation, proven with a prototype.
- Workshops - Hands-on sessions for engineering teams. Build, break, and understand autonomous agents in a weekend.
- Retainer - Ongoing experimentation for companies that want continuous answers.
Want to discuss your situation? Connect with me on LinkedIn to start a conversation.
Background
I'm a seasoned entrepreneur based in Paris with a passion for both hardware and software that spans over two decades. My journey began in 1995 when I stumbled upon a BASIC programming book for MO5, which soon evolved into a fascination with PC hardware when I got my first computer in 1996 (I had to disassemble it to understand how it worked).
In the late 90s, I moved quickly away from Quick Basic on to C and C++. I started developing small 3D games using OpenGL in 2002. I've always been intrigued by system-level programming and the concept of threads, which led me to explore Linux in depth, including building it from scratch in 2005 with LFS.
In 2003, I started assembling PCs at a local computer store in Grenoble, and by 2004, I was teaching computer basics to retirees. Graduating from ENSERG in 2007, I started working with microcontrollers and embedded Linux at Freescale.
I co-founded and lead a software development agency of 15 employees for 10 years. There I helped tailor software solutions that can run on-prem in secure environments for the broadcast industry with Orange (Globecast) and the energy sector with EDF.
Now I'm helping companies figure out whether AI agents can actually work for them - not through slides and vendor demos, but through working prototypes run long enough to see what breaks.
Because the only way to know if AI agents work is to actually run them.