Inserm U1219 Bordeaux Population Health · AHeaD team · University of Bordeaux
Where to place a nurse, how to order a queue, whether to open one more bed: these decisions are made today with no way of testing their effects other than on real patients. My PhD builds the model that lets you try them first — a simulator of the emergency department, calibrated on 225,766 real visits.
I come from theoretical physics and mathematics: an undergraduate physics degree at Orsay, a master's at ENS Ulm, then PDEs and probability at Bordeaux. An emergency department is, to me, a dynamical system like any other — except that its trajectories are patients, and getting it wrong costs hours of waiting.
An emergency department in miniature: seven areas, a finite number of places, and patients who only give theirs up once they find one further along. That single rule is what pushes the jam back towards the entrance. Play with the sliders.
The core of my PhD. An emergency department is a queueing network coupled to finite resources, clinical decisions and a saturated hospital downstream.
No flow model survives false data. The traces a care pathway leaves behind are produced in the course of clinical work, not for research: misordered events, impossible durations, duplicates, broken trajectories.
A model trained on a department’s history also learns its biases. With the team, I study the detection of human bias in triage — gender, age, socioeconomic status — using language models, and the use of serious games to make those biases perceptible to clinicians in training..
Automated assessment of pedestrian crossing safety across France, pairing aerial and street-level imagery to score nine risk factors. An applied project outside the hospital, sharing the same logic: producing at scale a measurement that human expertise cannot deliver across a whole country..
IEEE Journal of Biomedical and Health Informatics · 29(11):8484–8496
Transformer-based reconstruction of emergency activity logs at Bordeaux University Hospital, evaluated against an expert-annotated gold standard.
JMIR AI · 4:e57828
Reliable deidentification of emergency records using language models compact enough to be deployed on the hospital's own servers.
Heliyon · 11(4):e42428
Automated scoring of nine risk factors on French pedestrian crossings, from paired aerial and street-level imagery.
Published corrigendum Heliyon 11(14):e43902