Paul Michel
Research Scientist at DeepMind
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolPaul Michel is a research scientist with 11 years of experience working at the intersection of neural language modeling and robustness, currently at DeepMind after holding the Laplace Postdoctoral Chair in Data Science at École Normale Supérieure. His work centers on building adaptable NLP models that handle distributional shift and changing environments, a theme traced from his PhD at CMU's Language Technologies Institute through internships at Facebook and DeepMind. He combines strong engineering chops—contributing backend improvements to pytorch/translate and documenting the DyNet toolkit—with rigorous research on model robustness and adaptation. Based in London, he blends academic depth and practical system-building, and has a track record of improving tooling for translation and ensembling that eases reproducible experimentation.
11 years of coding experience
1 year of employment as a software developer
Ecole polytechniqueMP*, MP* at Lycée Louis-le-Grand
English, Japanese, German
Github Skills (11)
machine-learning10
restructuredtext10
python10
documentation10
sphinx9
command-line-arguments9
nlp9
doxygen9
pytorch9
beam-search9
artificial-intelligence8
Programming languages (4)
Github contributions (5)
clab/dynet
Oct 2016 - Oct 2018
DyNet: The Dynamic Neural Network Toolkit
Role in this project:
Technical Writer
Technical WriterContributions:1 release, 343 commits, 142 PRs in 2 years
Contributions summary:Paul's contributions focused on documenting the DyNet library. The primary task involved creating documentation for the library, including adding details on how to use specific operations in the code. The user's work resulted in the creation of various documentation files. The contributions demonstrate a focus on explaining and clarifying the usage of the library, which indicates a role centered on technical communication.
dynetdynamic-neural-networkdeep-learningneural-networksmachine-learning
pytorch/translate
May 2018 - Sep 2018
Translate - a PyTorch Language Library
Role in this project:Back-end Developer
Back-end DeveloperContributions:29 commits, 10 PRs, 2 branches in 3 months
Contributions summary:Paul primarily contributed to the `pytorch/translate` repository by adding features, refactoring, and improving the codebase. They implemented functionality to print translations to a file, simplifying command-line arguments, and streamlining beam search code. Furthermore, the user introduced source ensembling capabilities, enhancing the model's flexibility, and added the option to load pretrained weights during training.
pytorchnlptranslationmachine-learningonnx
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.