Paul Novello

ML Research Engineer, DEEL Team, IRT Saint Exupery, Toulouse.

prof_pic_paul.jpg

Myself 10 years ago.

Still looking the same.

Hi, I am Paul, an ML Research Engineer interested in Safe/Trustworthy AI and Scientific ML, i.e. ML applied to scientific computing and physical simulations. While these two fields might appear poled apart, there are connections (believe me). If you’re not convinced, see my research mindmap below.

I did my PhD at CESTA, Bordeaux’s research center of the French Alternative Energies and Atomic Energy Commission (CEA); CMAP, the Applied Mathematics lab of Ecole Polytechnique; and INRIA Saclay, Platon team. I tried to accelerate CFD simulations using AI, which took me three full years to achieve.

After that, I looked for a job where I could criterion 1: live in the south of France, criterion 2: do ML research, criterion 3: in a topic not so far from my PhD, criterion 4: all that in a permanent position. I was lucky enough to join the DEEL project in 2022, which was ticking every box. DEEL is a research project initiated in Toulouse by IRT Saint Exupery in 2018, which became international in 2020 when the Canadian counterpart was created in Montreal, and part of Artificial and Natural Intelligence Toulouse Institute (ANITI), one of the four French institutes for AI research (3IA) launched in 2020. The research topics revolve around trustworthy AI: Explainable AI (XAI), Out-Of-Distribution (OOD) detection, Conformal Prediction, Robustness, Fairness,… At that time, the team was gaining momentum, with 5 papers accepted at NeuIPS (for a dozen core members). The year after I joined, we got 8!

I quickly took the lead in the research effort on Out-Of-Distribution (OOD) detection and soon decided to try the open-source development experience with the creation of Oodeel, following the footsteps of Xplique and Deel-lip developed earlier by my teammates. In the meantime, I am still pursuing more classical research work.

Check out my research mindmap and reach out if interested in any topic!

news

Feb 20, 2024 Check out my synthetic list of past paper acceptance
Feb 20, 2024 Check out my synthetic list of past oral communications
Sep 04, 2023 v0.1.0 of oodeel! (First public release on PyPI)
Mar 09, 2022 PhD defense
Feb 15, 2022 Started an ML research Engineer at IRT Saint Exupery, DEEL core team!

selected publications

  1. Making Sense of Dependence: Efficient Black-box Explanations Using Dependence Measure
    Paul Novello , Thomas Fel , and David Vigouroux
    Advances in Neural Information Processing Systems (Neurips 2022), 2022
  2. Accelerating hypersonic reentry simulations using deep learning-based hybridization (with guarantees)
    Paul Novello , Gaël Poëtte , David Lugato , and 2 more authors
    Journal of Computational Physics, 2024
  3. Goal-oriented sensitivity analysis of hyperparameters in deep learning
    Paul Novello , Gaël Poëtte , David Lugato , and 1 more author
    Journal of Scientific Computing, 2023