Jean Feydy
Machine Learning Researcher at Inria
Paris, Ile-de-France
Summary
🤩
Rockstar🎓
Top SchoolJean Feydy is a machine learning researcher based in Paris with nine years of experience bridging geometric methods, medical applications, and high-performance GPU computing. Trained at École Normale Supérieure de Cachan, he combines strong theoretical grounding with hands-on engineering, notably contributing to KeOps — an influential library for kernel operations on CPUs and GPUs — where he fixed CUDA and C++ compatibility issues and prepared kernels for batch processing. His work sits at the intersection of geometry and scalable autodiff, enabling memory-efficient computations for large-scale models. Comfortable in both research and back-end development, he brings a practical focus to making advanced math run reliably in production-grade GPU environments. An underappreciated strength is his ability to troubleshoot low-level compiler and kernel problems that unlock higher-level algorithmic improvements.
10 years of coding experience
École Normale Supérieure de Cachan
Github Skills (8)
cuda10
c-language10
cprogramming-language10
kernel-mode9
template-metaprogramming9
back-end-development9
kernel9
autodiff8
Programming languages (3)
Github contributions (5)
getkeops/keops
Oct 2017 - Aug 2022
KErnel OPerationS, on CPUs and GPUs, with autodiff and without memory overflows
Role in this project:
Back-end Developer
Back-end DeveloperContributions:2 releases, 9 reviews, 517 commits in 4 years 10 months
Contributions summary:Jean's contributions focused on resolving compilation errors and integrating code changes related to building and merging branches in the KeOps library. The user addressed compatibility issues with C++11 and the nvcc 7.5 compiler, as well as modifications in the autodiff files and the CUDA kernels. These changes include fixing bugs in the Gaussian kernel and preparing the codebase for batch processing.
autodiffkernel
jeanfeydy/geomloss
Feb 2019 - Aug 2022
Geometric loss functions between point clouds, images and volumes
Contributions:1 release, 2 reviews, 151 commits in 3 years 6 months
loss-functionspoint-cloud
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.