Research interests

My research concerns the design, development, and analysis of high performance parallel numerical algorithms.

My recent work has particularly focused on accelerating linear algebra computations by harnessing numerical approximations, such as low-rank compression, mixed precision algorithms, and randomization.

Publications

Please refer to my publications and talks pages for a list of my recent research activities. You can also check out my Google Scholar profile.

Open positions

We have multiple open positions at all levels (Master, PhD, postdoc, and permanent positions) available in our team. Please contact me if you are interested. See also here for currently available Master internships.

Teaching

2024/2025 Lectures

• Arithmétique flottante et analyse d'erreur (AFAE, Sorbonne Université): lectures available on Moodle
• Harnessing inexactness in scientific computing (ENS Lyon)

For lectures from previous years, please contact me.

Projects

Current projects

• MUMPS
• NumPEx
• InterFLOP
• MixHPC

Past projects

• SEISCOPE
• ICONIC
• STRUMPACK
• MAGMA

Supervision

Postdocs

• Yongseok Jang (2024–ongoing, with P. Jolivet): Krylov solvers with mixed precision and deflation.
• Antoine Jego (2024–ongoing, with P. Amestoy, J.-Y. L'Excellent, G. Pichon): mixed precision memory accessors for sparse direct solvers.

PhD students

• Karmijn Hoogveld (2024–ongoing, with A. Buttari): mixed precision randomized low-rank approximations.
• Tom Caruso (2024–ongoing, with P. Jolivet, F. Nataf, P.-H. Tournier): mixed precision domain decomposition preconditioners.
• Hugo Dorfsman (2023–ongoing, with A. Anciaux, T. Guignon, F. Jézéquel): mixed precision BiCGstab solvers.
• Dimitri Lesnoff (2022–ongoing, with J. Berthomieu and S. Graillat): modular computing on GPUs for polynomial systems.
• Matthieu Robeyns (2021–ongoing, with M. Baboulin and O. Kaya): mixed precision low-rank matrix and tensor computing.
• Sébastien Dubois (2022–2024, uncompleted, with L. Grigori, C. Content, E. Martin): block low-rank domain decomposition preconditioners for CFD.
• Matthieu Gerest (2020–2023, with F. Jézéquel, O. Boiteau, and Mumps Tech): mixed precision block low-rank sparse direct solvers.
• Bastien Vieublé (2019–2022, with A. Buttari): mixed precision iterative refinement for sparse linear systems.

Master students

• Nicolas Catoni (2024, with E. Agullo, L. Giraud, F. Jézéquel, P. Jolivet): adaptive precision Krylov solvers.
• Roméo Molina (2021); then PhD (2021–2024) with F. Jézéquel: adaptive precision sparse matrix–vector product.
• Atef Dorai (2021, with F. Lopez): iterative refinement on GPU tensor cores.

Awards

• Lanczos ECMI/EMS Prize for mathematical software (with P. Amestoy and J.-Y. L'Excellent)
• Médaille de bronze du CNRS 2024
• SIAM Activity Group on Linear Algebra Early Career Prize 2021
• Second Prize for the 20th IMA Leslie Fox Prize
• Honorable mention for the 2020 Householder Prize
• Prix de thèse Gilles Kahn 2018