A work accepted for publication in Journal of Cheminformatics!
New work on Proxy-informed Bayesian transfer learning with unknown sources, stay tuned for the preprint!
A work accepted at NeurIPS 2024 Workshop on Bayesian Decision-making and Uncertainty on Computation-Aware Robust Gaussian Processes
New preprint on Preferential Amortized Black-Box Optimization
New work on Heteroscedastic Preferential Bayesian Optimization with Kernel Density Based User Modelling, stay tuned for the preprint!
I am a postdoc at Inria Bordeaux, in the SISTM team. I work on the interplay between data-driven and mechanistic models for Systems Biology in a Bayesian framework, a field I like to call Biologically-Informed Machine Learning.
Before that, I was a postdoc at Aalto University, in the Probabilistic Machine Learning group, working with Samuel Kaski, Vikas Garg and Markus Heinonen. I was working on Bayesian Inference, Bayesian Optimization, and the integration of human feedback into ML algorithms, with an emphasis on Drug Design.
And prior to that, I did my PhD at Inria and Institut Curie, under the supervision of François Fages and Annabelle Ballesta. We developed methods at the boundary of Systems Biology and Machine Learning to infer mechanistic models from time series data. A major contribution of the manuscript was to apply such methods to investigate the impact of systemic regulators (e.g. temperature, nutrient exposure, hormones) on the cellular circadian clock, in an attempt to personalize Cancer Chronotherapies. The manuscript can be found here and the slides here.
I serve as a reviewer for journals and conferences in different domains:
For a few years now, I have been cultivating a passion for indoor bouldering, with a mild success thus far ðŸ«
PhD in Computer Science, 2022
Inria / Institut Curie / Ecole polytechnique
Master's Degree 2nd year - Random modelling, Finance and Data Science, 2018
Université de Paris
Master's Degree 1st year - Applied Mathematics, 2017
Université de Paris
Learning misspecified mechanistic models from heterogeneous patient data
Human-In-The-Loop Machine Learning - Application to de novo drug design
On learning mechanistic models from time-series data with application to cancer chronotherapies