Research

My lab work sits between neuroscience, human factors and machine learning. These are the four threads, with the projects that carry them.

Research themes

Multimodal AI and human factors

Learning from real behaviour: gaze, hand motion, posture and neurocognitive signals, recorded where people actually work.

HABIT

Human-centred AI for behaviour learning in industrial technologies

Industry 5.0 needs systems that understand the operators who run them. HABIT models cognitive load and well-being on a real 5.0 production line at Icam Strasbourg-Europe, with ICube and the Human-IST lab of the University of Fribourg.

SkyMind

Pilot performance in simulated flight

Integrated eye tracking, hand tracking and AI to understand pilot attention and workload in simulated flight conditions.

Eye-tracking capture at VivaTech 2025: numbered fixations joined by gaze lines across the hall — Rest your pointer on the photo to leave your own fixations.

VivaTech 2025 study

Attention in a crowded innovation hall

A self-experiment with a Pupil Labs eye tracker at Europe's largest tech event: where gaze lands, how we navigate overload, and why we leave exhausted.

Trustworthy and explainable AI

AI people can question. Explainability metrics for security models, and generative agents limited to specific, ethical tasks.

speGAI

Specialised generative AI agents

Generative agents built for one specific, ethical task instead of answering everything.

Explainable intrusion detection

With Amani Abou Rida and Pierre Parrend

Benchmarks and metrics for how well explanation models describe cyberattack detection on IoT data.

Graphs and complex systems

Data where relations matter as much as values: graph neural networks, dynamic graphs and graph-tabular fusion.

AI for automotive data

CIFRE PhD with Cat Data

Hybrid graph-tabular models to reconcile and validate vehicle data at scale: VIN records, maintenance histories, parts and manufacturers.

Anomaly detection on graphs

Static and dynamic graphs

Inductive node embeddings with graph convolutional networks to spot attacks in network traffic.

Vision for infrastructure and industry

Where my research started: detecting thin cracks in pavement images, now extended to industrial inspection and safety.

MIIRAG

Imaging AI for safe roads

Maintenance through imaging and AI for safer roads and better infrastructure governance.

Laser weld prediction

With Victor Hayot and Icam colleagues

Machine learning that predicts laser weld penetration on multi-material data.

Recent openings

Posted in 2026. Write to me to check whether a position is still open.

PhD: Human-centred AI for behaviour learning (HABIT)Icam, ICube and University of FribourgAsk if it is still open
CIFRE PhD: AI for automotive dataIcam Strasbourg-Europe and Cat DataAsk if it is still open