AMJAX
Algebraic multigrid solvers in JAX.
Biography
Hey there, I’m Victor, an applied ML researcher working in AI for science, currently at the Information and Network Dynamics Lab at the Ecole Polytechnique Fédérale de Lausanne (EPFL).
I develop methods for high-dimensional, spatio-temporal forecasting and inverse problems that remain robust under low signal-to-noise ratios and distribution shift, with application in biology and environmental sciences. I am particularly interested in hybrid modelling, integrating domain priors with deep learning and GPU-accelerated differentiable computing. I ship open-source libraries implementing these methods, lead funded research projects, and mentor students in applied machine learning and scientific computing.
Outside of work, I am a part-time alpinist, passionate about mountain adventures and writing. I also enjoy sailing and surfing occasionally. You can check out my alpine CV here.
Interests
Click a topic to browse related work.
Education
-
PhD in Environmental Sciences, 2023
ETH Zürich, Switzerland
-
MSc in Energy and Environmental Sciences, 2018
INSA Lyon, France
Open source software 🧑🏽💻
ML library
HybridDynamicModels.jl
A Julia library for building and training hybrid dynamic models that combine mechanistic and data-driven components.
ML model
MuScaRi
Multi-scale model for spatial biodiversity estimation.
Domain-centric library
jaxscape
A minimal JAX library for graph-based connectivity analysis at scale.
EcoEvoModelZoo.jl
A Julia package providing access to a collection of eco-evolutionary models.
EvoId.jl
A Julia package for simulating evolutionary individual-based models.
Posts
Publications
I also review for the Journal of Open Source Software, Ecology Letters, Ecography, Biodiversity and Conservation, and Methods in Ecology and Evolution.
Peer-reviewed
(2026).
Chaotic Slow Slip Events in New Zealand from Two Coupled Slip Patches: A Proof of Concept.
ARC Geophysical Research.
(2025).
Spatial biodiversity indicators and a composite index for conservation prioritization in Switzerland.
bioRxiv.
Accepted in Ecology.
(2025).
A calibration framework to improve mechanistic forecasts with hybrid dynamic models.
Methods in Ecology and Evolution.
(2025).
Differentiable Programming for Differential Equations: A Review.
arXiv.
Accepted at SIAM Review.
(2025).
Species loss in key habitats accelerates regional food web disruption.
Communications Biology.
(2025).
Temporal horizons in forecasting: a performance-learnability trade-off.
Transactions on Machine Learning Research.
(2023).
Paleoenvironments shaped the exchange of terrestrial vertebrates across Wallace’s Line.
Science.
(2023).
Ancient sedimentary DNA to forecast trajectories of ecosystem under climate change.
Philosophical Transactions of the Royal Society B.
(2022).
Eco-evolutionary model on spatial graphs reveals how habitat structure affects phenotypic differentiation.
Communications Biology.
Books
(2024).
Accelerating renewable energy development while enhancing biodiversity protection in Switzerland.
RE-BD AR2024.
Contributing author of chapters 2 and 5.
Conference papers
(2026).
Biodiversity Profile Estimation with Earth Observation.
Accepted at the International Association for Pattern Recognition Workshop on Pattern Recognition in Remote Sensing 2026.
(2018).
Episodic mineralising fluid injection through chemical shear zones.
Australian Society of Exploration Geophysicists Extended Abstracts.
Preprints
(2026).
A Unified Framework for Prioritizing Habitat and Connectivity Conservation through Analytical Sensitivity.
bioRxiv.
Revision requested from Landscape Ecology.
(2026).
Clustering biomes from space: from pixels to foundation models.
SSRN.
Submitted to Remote Sensing of Environment.
(2025).
Multi-scale species richness estimation with deep learning.
arXiv.
Revision requested from Nature Communications.
Selected funding
EcoDynCast
Forecasting alien invasive species range dynamics with GNNs.
100,000
CHF
A course on reproducible research, data pipelines, and scientific computing
Funding for the CORDS course, coordinated by Mauro Werder.
50,000
CHF
yDiv workshop funding
Funding for a three-day Julia workshop for biodiversity research.
10,000
EUR
WSL Biodiversity Center Innovative Workshop Grant
Funding for a Julia workshop on modelling and data analysis in biodiversity and earth sciences.
10,000
CHF
Teaching
Students
Current
Capucine Lechartre
PhD co-supervision
A mechanistic approach to biome transitions across space and time
Ghali Elouahdani
Semester project
Fanny Missier
Semester project
Jeffrey Zweidler
Master thesis
Forecasting invasive species range expansion using ecologically-informed neural networks
Co-supervision with Swiss Data Science Center
Moritz Dieing
Master thesis
Attention-based deep multiple instance learning for species richness prediction
Co-supervision with Swiss Data Science Center
Alumni
Cecilia Valenzuela Agui
Taste of research internship
Nicolas Demolin
Research internship
Teaching
2024
- 2-day Course On Reproducible Research, Data Pipelines, and Scientific Computing
- Landscape Modelling of Biodiversity: From Global Changes to Conservation, mentoring the project Modelling the spread dynamics of invasive alien species, ETH Zürich, D-USYS
2023
- Environmental Systems Data Science, in charge of the unit Supervised Deep Learning - Application, ETH Zürich, D-USYS
- 2 day workshop on Julia for modelling and data analysis in biodiversity and earth sciences