Home

Training students who understand their tools, and designing tools that leave nobody out.

Hello!

Welcome to my (semi-)professional website.

Engineering graduate of ENSIMAG (2012), PhD in computer science (Université Toulouse 3, 2017; thesis on text entry on mobile devices for visually impaired users), associate researcher with the SIG team at IRIT, I am an Associate Professor in computer science at 3iL Ingénieurs, on the Rodez campus (in Aveyron, France!), where I have taught since 2020. I cover microservice architecture, advanced development, object-oriented programming, software architectures and software engineering.

My first publication dates from 2013, the most recent from 2026: I did not arrive on these topics with the wave of conversational agents. What has not changed in between is the method. A protocol written before the experiment, published datasets, figures anyone can reproduce, and work submitted to peer review.

My research is organised around three areas:

Ontology alignment and explainability. Knowledge fusion and the OAEI campaigns, with one central question: how to explain, verify and make interpretable a result produced by an automated system.
Accessibility, disability and AI systems. Text entry for visually impaired users, the subject of my PhD, and then how generative AI handles disability: which guardrails, applied to whom, with what asymmetries.
Generative AI and engineering education. What students and teachers actually do with it, measured rather than assumed, and what that entails in terms of responsibility when you train people to build systems.

More on these areas · The full list of publications, with DOIs, PDFs and BibTeX entries.

One example of what that looks like in practice: Gemini, ChatGPT, Claude and disability, a study I ran myself, with its paired prompts, its datasets and its reproducible figures.

The word AI is used to sell a great deal these days, and I read those claims with suspicion. I use these tools daily and I teach them; what I take issue with are figures quoted without a protocol, self-declared expertise, and promises nobody is bound to keep. Several pieces here take that on: What AI is Not, The tool is not for sale, and Two hours in Baybay, a talk whose subtitle already said the essential.

I place particular importance on simple, accessible, and clear computing, with a fondness for sober, retro, and slow-tech approaches. What that amounts to here, what has been tested and what remains to be fixed: see the accessibility page.

Since my surgery in January 2025, I am officially a cyborg, which makes me especially attentive to (and personally concerned by) issues of accessibility, disability, and inclusion in digital tools.