Mathieu Blondel
Staff Research Scientist, Manager at Google
Paris, Ile-de-France
Summary
🤩
Rockstar🎓
Top SchoolMathieu Blondel is a Staff Research Scientist and manager at Google with 20 years of experience in machine learning, natural language processing and scientific computing. He combines deep academic training (Doctor of Engineering in Machine Learning from Kobe University) with a strong engineering bent, contributing significant optimization and data-loading implementations to open-source projects such as JAXopt and scikit-learn-contrib/lightning. At Google he has progressed from senior research scientist to staff and manager, bridging research and production by implementing hardware-accelerated optimizers (FISTA with implicit differentiation) and practical dataset utilities (MNIST loaders and non-contiguous data handlers). Based in Paris, he brings a rare mix of algorithmic rigor and hands-on backend engineering that improves both model performance and developer ergonomics.
20 years of coding experience
12 years of employment as a software developer
Engineering diploma, Telecom, Engineering diploma, Telecom at Télécom LilleDoctor of Engineering, Machine Learning, Doctor of Engineering, Machine Learning at Kobe University
French, English, Japanese, Chinese
Stackoverflow
280reputation
75kreached
3answers
2questions
Github Skills (21)
algorithm10
python10
optimizers10
data-set10
machine-learning10
datasets10
optimisation10
data-handling10
jax10
optimization10
scikit9
scikit-learn9
linear-algebra9
differentiable-programming9
backend8
Programming languages (7)
Github contributions (5)
google/jaxopt
Feb 2021 - Jan 2023
Hardware accelerated, batchable and differentiable optimizers in JAX.
Role in this project:
ML Engineer
ML EngineerContributions:13 releases, 314 reviews, 126 commits in 1 year 11 months
Contributions summary:Mathieu implemented and tested FISTA, a hardware-accelerated optimizer within the JAXopt library. They also added support for implicit differentiation, and incorporated a loop structure to increase efficiency of the algorithm. Furthermore, they enhanced the library by incorporating a multi-class logistic loss function and a proximal operator for l1 regularization. The changes focused on optimization algorithms and related functionalities.
pytorchdifferentiabledifferentiable-programmingautomatic-differentiationdeep-learning
scikit-learn-contrib/lightning
Jan 2012 - May 2021
Large-scale linear classification, regression and ranking in Python
Role in this project:Back-end Developer
Back-end DeveloperContributions:14 reviews, 557 commits, 54 PRs in 9 years 5 months
Contributions summary:Mathieu implemented core functionalities for dataset loading and processing within the project. They added dataset loaders for various datasets, including MNIST, and integrated them into the project's structure. The user also added specific methods for handling non-contiguous data structures. These changes expanded the project's capabilities by providing access to data loading utilities and improving the overall usability for machine learning tasks.
regressionpythonclassification-algorithmlogistic-regressiondata-science
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.