Mission & Identity
Building AI systems that operate at production scale — and asking, at every step, whether they serve humanity or merely optimise for it.
Combining rigorous applied science with philosophical discipline. The best engineering decisions require slowing down — the friction is the point.
Formed between Casablanca and Paris. Disciplined by the tatami. Drawn equally to algorithms, literature, and the open sea.
My forthcoming essay asks a simple, uncomfortable question: at what point does relying on intelligent machines stop being convenient and start becoming cognitive atrophy? Grounded in years of building the systems themselves, La paresse de penser is not a warning — it is an invitation to think more carefully about what we choose to stop thinking about.
About
I am a Senior Applied Scientist at Microsoft, based in Seattle, where I architect NLP models and AI systems within the Microsoft 365 ecosystem — work that has produced 9 patents spanning intelligent personalisation, behavioural anomaly detection, and human well-being in the workplace.
I trained as an engineer at Télécom Paris, where I first understood that the most interesting problems are never purely technical. I continue to work at that boundary: I lecture at EPITA Paris on advanced NLP and large language model architectures, and I mentor Applied Scientists across Microsoft on machine learning rigour and privacy-preserving AI.
Outside the lab, discipline comes from the tatami. I hold a black belt in Judo and am a former two-time national champion in Karate Jutsu. I surf the waters of the Pacific Northwest. And I write — most recently a book on why the age of intelligent machines demands, more than ever, that we think for ourselves.
Intellectual Property & Research
Methodology for intelligent information streamlining in collaborative virtual spaces.
Adaptive systems that learn user behaviour to optimise high-performance workflows.
A linguistic AI model fostering inclusivity through accurate global name recognition.
Personalised memory systems for computing interfaces to enhance human recall.
AI-driven system to detect and enhance the quality of virtual collaboration.
Contextual AI for generating appropriate, personalised out-of-office communications.
Next-generation NLP for processing feedback while maintaining strict privacy compliance.
Detecting and mitigating digital fatigue through AI-driven behavioural analysis.
Behavioural AI for real-time ergonomic and posture guidance in computing environments.
Exploring hyperbolic geometry for better hierarchical data representation in machine learning.
From the Blog
The Laziness of Thinking
An introduction to my upcoming book — and to the question of what we lose when intelligent machines think for us.
Read →What LLMs Taught Me About Human Mental Laziness
The mechanics of next-token prediction turn out to be a remarkably good model of how humans cut corners.
Read →