About me
Hi, my name is François! I solve challenging scientific problems with AI. I am interested in many topics, including generative modeling, inverse problems, and physics emulation, both from an application and theoretical perspective. I hold a PhD in Machine Learning from the University of Liège in Belgium, advised by Prof. Gilles Louppe. I currently work as a Research Scientist at the Ellison Institute of Technology.
I am also passionate about software development and open-source software. I regularly contribute to open-source projects and have published several libraries with thousands of daily users such as Zuko, PIQA and Inox.
For more information, take a look at my resume.
Publications
Here follows a list of selected publications. For the full list, see my scholar page.
-
Walrus: A Cross-domain Foundation Model for Continuum Dynamics
McCabe, Mukhopadhyay, Marwah et al. in International Conference on Machine Learning (2026)
-
Generative Modeling in Large-scale Dynamical Systems
Rozet (2026)
-
Training-Free Bayesian Filtering with Generative Emulators
Savary, Rozet and Louppe in International Conference on Machine Learning (2026)
-
Appa: Bending Weather Dynamics with Latent Diffusion Models for Global Data Assimilation
Andry, Rozet, Lewin et al. (2025)
-
Lost in Latent Space: An Empirical Study of Latent Diffusion Models for Physics Emulation
Rozet, Ohana, McCabe et al. in Advances in Neural Information Processing Systems (2025)
-
The Well: a Large-Scale Collection of Diverse Physics Simulations for Machine Learning
Ohana, McCabe, Meyer et al. in Advances in Neural Information Processing Systems (2024)
-
Learning Diffusion Priors from Observations by Expectation Maximization
Rozet, Andry, Lanusse and Louppe in Advances in Neural Information Processing Systems (2024)
-
Score-based Data Assimilation
Rozet and Louppe in Advances in Neural Information Processing Systems (2023)
-
Towards Reliable Simulation-Based Inference with Balanced Neural Ratio Estimation
Delaunoy, Hermans, Rozet, Wehenkel and Louppe in Advances in Neural Information Processing Systems (2022)
-
A Trust Crisis In Simulation-Based Inference? Your Posterior Approximations Can Be Unfaithful
Hermans, Delaunoy, Rozet, Wehenkel and Louppe in Transactions on Machine Learning Research (2022)