Founder.
Builder.

Kanop is the geospatial intelligence partner for nature. We combine AI, satellite imagery and scientific expertise to estimate above-ground biomass, track biodiversity, and deliver rigorous MRV for nature-based solutions and supply chains. Operating globally, across forests and ecosystems on every continent. Built from scratch in 2021, now a team of engineers and scientists working at continental scale.

Built deep learning systems for oncology that ended up in clinical use. Automated segmentation of 80+ organs at risk, multi-modal image registration across CT and MRI, dose planning for radiotherapy. FDA and CE certified. Now used in roughly 20% of radiotherapy treatments in France. Did it all in MXNet, before PyTorch won the war.

Started writing ML models in R. At Shippeo, built arrival time prediction that outperformed Google Maps and HERE APIs for freight-specific logistics. At Lucky Cart, modelled promotional ROI in a market where 59% of promotions lose money. A different era and a good one to have started in.

Started with pure mathematics at EPFL, the kind that makes everything else feel tractable. Moved into machine learning when the term still needed explaining, graduating top of class at CentraleSupélec's MSc in AI. Spent time across three countries, three research cultures, and enough linear algebra to last a lifetime. Wrote a public blog documenting what I was learning, and a section on climate change that turned out to be a preview of the next decade.

Extending InSAR2InSAR to Sentinel-1 Data · IEEE Geoscience and Remote Sensing Letters · 2025

InSAR2InSAR: A Self-Supervised Method for InSAR Parameters Estimation · EUSIPCO · 2024