Grégory Chatel
Lead R&D And Data Scientist
Gagny, Île-de-France, France
Summary
👤
Senior🎓
Top SchoolGrégory Chatel is a Lead R&D and Data Scientist with a PhD in computer science and 11 years of experience building deep learning solutions, currently applying AI to environmental problems via satellite imagery at DISAITEK. He combines academic rigor—teaching deep learning at Université Gustave Eiffel and a doctorate in combinatorial algebra—with practical industry impact from roles at Société Générale and Intel’s Software Innovator program. A hands-on contributor to major open-source NLP projects (notably enhancements to Hugging Face’s transformers and a PyTorch OpenAI transformer implementation), he focuses on adapting transformer architectures and dataset integrations for real-world tasks like multiple-choice reasoning and similarity. Known for bridging theory and production, he brings a rare mix of mathematical depth, model-level engineering, and applied R&D driven by environmental and regulatory use cases.
11 years of coding experience
1 year of employment as a software developer
Doctorat, Informatique, combinatoire algébrique, Doctorat, Informatique, combinatoire algébrique at Université Paris-Est Marne-la-Vallée
English
Github Skills (15)
transformers10
transformer-models10
pytorch10
machine-learning10
nlp10
language-model10
deep-learning10
python10
bert10
natural-language-processing10
pre-trained-model9
modeling9
modello9
nlpjs9
modeler9
Programming languages (4)
Github contributions (5)
huggingface/pytorch-openai-transformer-lm
Jun 2018 - Aug 2018
🐥A PyTorch implementation of OpenAI's finetuned transformer language model with a script to import the weights pre-trained by OpenAI
Role in this project:
ML Engineer
ML EngineerContributions:25 commits, 11 PRs, 27 comments in 2 months
Contributions summary:Grégory primarily focused on refining and extending the functionality of a PyTorch-based transformer language model. Their contributions included refactoring existing code for clarity, introducing new head modules for tasks such as classification and similarity, and modifying the model's architecture to accommodate different task types. Furthermore, the user made adjustments to the data encoding and loss computation processes. The changes aimed to improve flexibility and support for various downstream applications.
pytorchnlptransformerslanguage-modeltransformer-models
huggingface/transformers
Dec 2018 - Jul 2019
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:ML Engineer
ML EngineerContributions:23 commits, 7 PRs, 44 comments in 7 months
Contributions summary:Grégory primarily focused on developing and refining code related to the SWAG (Situations With Actions and Goals) dataset, a multiple-choice task used for evaluating language models. Their contributions included defining the `SwagExample` class, implementing code to read the dataset, creating the `convert_examples_to_features` function, and integrating it with a `BertForMultipleChoice` model. The user also fixed commentary and improved the code structure for clarity and readability. This work involved integrating BERT models with a new multiple-choice dataset.
pythonbertspeech-recognitionstate-of-the-artflax
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.