Foundation models for biological discovery
I am a PhD student at Mila and Polytechnique Montréal, studying foundation models for multi-omics data, with a focus on protein language models. My research explores what these models learn and how their representations can support biological discovery. I work under the supervision of Prof. Sarath Chandar and Prof. Quentin Fournier.
Publications & preprints
pLM representations unlock metagenomic space beyond homology
Investigated how pretraining data composition shapes protein language models, revealing a trade-off between calibration to evolutionary distributions and structural prediction. Introduced RED, a scalable metric derived from residue-embedding geometry highly correlated with pseudo-perplexity and a task evaluating alignment to natural evolutionary variation. Developed a method to retrieve candidates from large metagenomic databases and experimentally validated their in vivo expression.
A biological-response compound representation allows chemical perturbation prediction across cell lines
New compound representation method for the prediction of transcriptomic perturbations in cancer cell lines. Biological activity enables successful cross-cell line predictions, where chemical representations derived from molecular foundation models fail.
NeoBERT: A Next-Generation BERT
NeoBERT is a modernized BERT-like model with improved data, training strategy and architecture. It achieved top-of-the-leaderboard scores in its category on MTEB, a widely used sentence embedding benchmark.
Talks & posters
pLM representations unlock metagenomic space beyond homology
6th Applied Biocatalysis & Enzyme Engineering conference
Related paper: pLM representations unlock metagenomic space beyond homology
Teaching & service
Royal Military College of Canada
Taught a seminar on training on a HPC environment, pretraining, and distributed learning.
Machine Learning (INF8245AE)
Teaching assistant for Prof. Sarath Chandar.
CRL Summer School Program
Taught introductory computer science courses at Mila to high school students from underrepresented backgrounds in July 2025 and July 2026.
CRL Application Assistance Program
Founded and led a yearly program at Mila mentoring students from underrepresented communities applying to graduate studies in AI, from September 2023 to September 2025.
Education
September 2023 – present
Polytechnique Montréal
PhD
Fast-tracked from the research master’s program in January 2025. Coursework: Machine Learning, Representation Learning, Probabilistic Graphical Models, and Dynamical Systems.
September 2022 – August 2023
École 42
Software engineering program
Project-based learning and peer-to-peer collaboration.
September 2019 – August 2022
École Polytechnique
Diplôme du cycle ingénieur polytechnicien (X2019)
Majors in Computer Science and Applied Mathematics. Minors in Biology, Physics, and Environmental Sciences.
September 2017 – August 2019
Lycée Saint-Louis
Preparatory classes, MPSI/MP*
About me
I grew up in French Polynesia, a group of small islands in the Pacific Ocean, and spent most of my studies in Paris before moving to Montréal in 2023.