Current members

• Laurynas Varnas — PhD student (2026–) at Toulouse-INP/ANITI; Exascale training algorithms for next-generation weather prediction with graph neural networks. Co-supervised with S. Gratton, J. Herrmann (Toulouse-INP/ANITI).
• Aymane Kssim — PhD student (2025–) at Toulouse-INP/ANITI/UNIGE; Parallel training algorithms for scientific machine learning. Co-supervised with S. Gratton (Toulouse-INP/ANITI).

Alumni

PhD students

• Dr. Samuel Cruz — PhD (USI, 2022–2025), Domain-decomposition methods for machine learning, together with R. Krause.

Master theses

• Martin Masson — M2 research stage (Paris Saclay/Toulouse-INP, 2026), Optimization for meta-learning. Université Paris-Saclay.
• Laurynas Varnas — MSc thesis (USI, 2024–2026), Decomposing graph neural networks, together with A. Heinlein.
• Mahmoud Aharmouch — M2 stage (Paul Sabatier University, 2025), AI-enhanced nonlinear iterative methods.
• Marc Salvadó — MSc thesis (USI, 2023–2024), Multilevel approaches to enhance the training of transformer models, together with R. Krause.
• Andrea Angino — MSc thesis (USI/Insubria, 2021–2022), Knight descent – a parallel stochastic method for nonlinear optimization problems, together with R. Krause and M. Donatelli.
• Samuel Cruz — MSc thesis (USI, 2019–2020), Learning multilevel hierarchies, together with R. Krause.
• Vanessa Braglia — MSc thesis (USI, 2019–2020), Multilevel training for neural networks, together with R. Krause.

Bachelor theses

• Stefano Gonçalves — BSc thesis (USI, 2021–2022), Implementation of a hybrid data-parallel algorithm for neural network training with reduced communication targeted to GPU-based supercomputers, together with R. Krause.
• Filippo Cesana — BSc thesis (USI, 2020–2021), Python front-end for Utopia with algorithmic implementations related to financial machine learning, together with R. Krause and P. Zulian.

Interns and student assistants

• Beya Hachicha — M1 stage (Toulouse-INP, 2025), Extreme learning machines.
• Maxime Hanus — M1 stage (Toulouse-INP/ESILV, 2025), Operator learning for unstructured geometries.
• Marc Salvadó — USI (2022–2023), Layer-parallel training of large language models, together with J. Schroder (University of New Mexico) and E. Cyr (Sandia National Laboratories).
• Francesco Lacommare — ETH Zurich (2021), Multilevel variant of the Adam optimizer.
• Filippo Cesana — USI (2020–2021), Python interface for UTOPIA, together with P. Zulian.
• Dylan Ramelli — USI (2020–2021), xSDK integration for UTOPIA, together with P. Zulian.
• Nicholas Robertson — USI/EPFL (2020–2021), Domain decomposition and machine learning.
• Lisa Gaedke-Merzhäuser — Freie Universität Berlin/USI (2019–2020), Multilevel training of deep residual networks.
• Samuel Cruz — USI (2019–2020), Learning multigrid transfer operators using reinforcement learning.
• Vanessa Braglia — USI (2019–2020), Multilevel variance reduction methods.
• Eric Botter — USI (2016), Continuous integration using CDash.