Current position

I am Professor in applied mathematics (speciality statistics) in Université Côte d’Azur since 2022. I am member of the of probabilities and staticitic team of the J.A. Dieudonné Laborary, and of the MAASAI team of the Inria Center of Université Côte d’Azur.

I hold a Chair in Artificial Intelligence from the 3IA Côte d’Azur insititute in the Core Elements of AI axis of the institute. My chair is entitled Finding structure in heterogeneous data.

I am deputy scientific director of the EFELIA Côte d’Azur project, a five-year ANR CMA project to massify AI training in all domains of Université Côte d’Azur.

Previous position

Previously, I was Assistant Professor in applied mathematics (speciality statistics) in the STID department of the IUT C of the University of Lille from 2010 to 2022. I was a member of the ULR 2694 METRICS “Health Technology Assessment and Medical Practice” and of the MODAL “MOdel for Data Analysis and Learning” team of Inria Lille Nord Europe.

Research activities

My research work is in the field of statistics and more particularly in the field of classification (supervised, unsupervised and semi-supervised), approached from the perspective of probabilistic models (model-based clustering). In the continuity of my thesis work, a significant part of my research work deals with clustering based on mixture models. In this context, I have contributed to the proposal of models, in particular models allowing to take into account dependencies between variables (categorical, mixed) as well as for the clustering of functional data. I have also contributed to the issue of multiple partition clustering (i.e. when several class latent variables are considered) for which we have proposed two models and their associated estimation procedures. Finally, I have also contributed to the general study of mixture models in clustering through the questions of taking into account missing values and the behaviour of the EM algorithm in these cases, the problem of label-switching in Bayesian statistics, and the proposal of an approach to visualize the results of any mixture model.

I have fed my methodological research with applied problems. On the one hand in the medical field, such as the implementation of mixture models for the classification of patient pathways in hospitals, the prediction of the number of problems in the field of usability of medical devices, or the use of high-dimensional regression models in proteomics. On the other hand, I have been interested in the industrial and retail field through the proposal of automatic discretization models in \emph{credit scoring} and the use of machine learning in the aviation field.