Rémi Louf
Paris, Ile-de-France
Summary
🤩
Rockstar🎓
Top SchoolRémi Louf is a researcher-engineer with 12 years of experience blending probabilistic modeling, generative models and Bayesian inference into production ML systems; since 2018 he has led scientific efforts at Sounds as Chief Science Officer. Trained at top French institutions and Oxford with a PhD in statistical physics, he brings deep theoretical grounding to practical problems in causal and casual inference, A/B testing and sequence generation. His open-source contributions span high-impact projects—improving graph algorithms in Julia, refactoring BERT for generation in Hugging Face Transformers, and advancing JAX-backed probabilistic primitives in PyTensor/Aesara—showing fluency from mathematics to performant backend code. Known for fixing subtle numerical and documentation issues alike, he pairs rigorous academic thinking with pragmatic engineering that ships.
12 years of coding experience
Élève normalien, Fundamental Physics, Élève normalien, Fundamental Physics at École normale supérieure de CachanMaster of Studies, Philosophy of Physics, Master of Studies, Philosophy of Physics at University of OxfordMaster's degree, Quantum Physics, Master's degree, Quantum Physics at Ecole normale supérieureDoctor of Philosophy (PhD), Statistical Physics, Doctor of Philosophy (PhD), Statistical Physics at Institut de Physique Théorique
French, English
Github Skills (39)
algorithm10
algorithms10
pytorch10
probabilistic-programming10
py10
python10
pre-trained-model10
architecture10
testing10
machine-learning10
data-structure10
refactor10
automatic-differentiation10
compiler-design10
bayesian10
Programming languages (14)
Github contributions (5)
dottxt-ai/outlines
May 2023 - Apr 2025
Structured Text Generation
Role in this project:
Back-end Developer & Test Automation Engineer
Back-end Developer & Test Automation EngineerContributions:56 releases, 384 reviews, 534 PRs in 1 year 10 months
Contributions summary:Rémi's commits focused on adding and testing compiler functionality within the "outlines" project. They implemented tests for the compiler module and created test scripts to verify the functionality of the structured text generation features, including testing for string variables and script compilation. The user's work directly addressed testing and validation of the core features of the library which aligns with the project's goal of structured text generation. The commits also extended the library's functionality to include the Mako templating language.
cfggenerative-aijsonllmsprompt-engineering
aesara-devs/aesara
Jun 2021 - Dec 2022
Aesara is a Python library for defining, optimizing, and efficiently evaluating mathematical expressions involving multi-dimensional arrays.
Role in this project:Back-end Developer
Back-end DeveloperContributions:2 releases, 82 reviews, 100 commits in 1 year 6 months
Contributions summary:Rémi made several commits focused on refactoring and optimizing code related to mathematical operations within the Aesara library. They renamed functions, updated code within `math_opt.py` to improve simplification rules and lifting behavior, and modified `basic.py` and `math.py` to adjust built-in functions. Additionally, they added, updated, and improved documentation.
multi-dimensionalpythonmathematical-expressionsautomatic-differentiationoptimizing-compiler
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.