Louis Béthune
Research Software Engineer at Apple
Paris, Ile-de-France
Summary
🤩
Rockstar🎓
Top SchoolLouis Béthune is a Research Software Engineer based in Paris with nine years of experience at the intersection of deep learning, reinforcement learning and graph-based methods. He holds a master's from ENS Lyon and completed a PhD focused on Lipschitz constraints for deep learning, translating theoretical insight into practical model improvements. Louis has research and industry experience across top labs and companies—Google, Mila, Inria—and now Apple, where he applies rigorous experimentation to scalable ML systems. He is an active contributor to JAXopt, implementing advanced optimization and fixed-point methods (e.g., Anderson Acceleration) and ensuring correctness via gradient-aware tests. Known for probing why models work as much as for building them, he blends mathematical rigor with hands-on software engineering to push optimizer and RL tooling forward. Colleagues describe him as someone who bridges deep theory and production-quality code, often surfacing subtle stability and convergence improvements that others miss.
9 years of coding experience
MPSI/MP, MPSI/MP at Lycée Henri WallonMaster's degree, Computer Science, Master's degree, Computer Science at École normale supérieure de Lyon
Stackoverflow
13reputation
225reached
0answers
1question
Github Skills (11)
algorithm10
differentiable-programming10
jax10
optimisation10
optimizers10
python10
unit-test10
optimization10
machine-learning9
deep-learning6
eigen6
Programming languages (5)
Github contributions (5)
google/jaxopt
Sep 2021 - Nov 2022
Hardware accelerated, batchable and differentiable optimizers in JAX.
Role in this project:
ML Engineer
ML EngineerContributions:29 reviews, 37 commits, 24 PRs in 1 year 2 months
Contributions summary:Louis primarily contributes to the development and testing of optimization algorithms within the JAXopt library. Their work involves implementing and testing Anderson Acceleration and related fixed-point iteration methods. The commits demonstrate the addition of new algorithms, along with associated unit tests to verify their correctness, including gradient checks. The user's efforts appear focused on expanding the library's capabilities in optimization and fixed-point methods.
pytorchdifferentiabledifferentiable-programmingautomatic-differentiationdeep-learning
Algue-Rythme/jaxopt
Oct 2021 - Jun 2022
Hardware accelerated, batchable and differentiable optimizers in JAX.
Contributions:10 commits, 156 pushes, 64 branches in 7 months
pytorchdifferentiableautomatic-differentiationoptimizationaccelerated
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.