Quentin Gallouédec
Research Engineer at Hugging Face
Montreal, Quebec, Canada
Summary
🤩
Rockstar🎓
Top SchoolQuentin Gallouédec is a research engineer at Hugging Face with seven years’ experience applying reinforcement learning and transformer-based methods to production-grade ML tooling. He is the lead maintainer on TRL (training transformer language models with RL) and has made substantive back-end contributions to high-profile projects like Hugging Face Transformers and Stable Baselines3, improving trainers, replay buffers, and multi-device support. Quentin holds an engineering degree and a PhD in Machine Learning from École Centrale de Lyon and blends deep research rigor with practical engineering—refactoring dataset handling, loss logic, and memory management for real-world workloads. His background includes robotics and real-time control research (a 3D-printed robotic arm project) and hands-on systems experience from implementing GRPO rewards to fixing subtle prediction-type bugs. He also serves as a reservist innovation officer in the Paris fire brigade, a role that informs his pragmatic approach to resilient system design. Fluent across PyTorch, RL algorithm implementations, and large-model training pipelines, he thrives on making complex research reproducible and production-ready.
7 years of coding experience
Master’s degree, Electronics, Electrical Energy and Automation / Electronics of embedded systems, Master’s degree, Electronics, Electrical Energy and Automation / Electronics of embedded systems at CPE LyonClasses préparatoires aux grandes écoles (CPGE), Classes préparatoires aux grandes écoles (CPGE) at Lycée Clemenceau - NantesPhilosophiæ doctor (Ph.D.), Machine Learning, Philosophiæ doctor (Ph.D.), Machine Learning at École Centrale de Lyon
Github Skills (25)
algorithm10
transformers10
pytorch10
python10
machine-learning10
reinforcement-learning10
natural-language-processing10
nlp10
data-structure9
algorithms9
gymnasium9
openai-gym9
language-model9
huggingface9
hyperparameter-optimization9
Programming languages (11)
Github contributions (5)
huggingface/trl
Jan 2023 - Jul 2026
Train transformer language models with reinforcement learning.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:24 releases, 1944 reviews, 1133 PRs in 3 years 7 months
Contributions summary:Quentin primarily focused on improving the DPO Trainer, addressing a typo in the DPOTrainer's warnings and refactoring parameters. They refactored the code to handle various dataset formats and updated the DPO code by truncating input sequences and padding. The user also made changes to the code related to handling different loss types, indicating proficiency in the core DPO training process.
language-modelreinforcement-learning
huggingface/open-r1
Jan 2025 - Sep 2025
Fully open reproduction of DeepSeek-R1
Role in this project:ML Engineer
ML EngineerContributions:87 reviews, 26 PRs, 31 pushes in 7 months
Contributions summary:Quentin primarily focused on developing and maintaining the GRPO script, a core component of the project. They implemented reward functions, including accuracy and format checks, and integrated the cosine scaled reward function. Furthermore, the user contributed to fixing bugs in the GRPO script and refactoring the code for compatibility. Their work included code modifications and updates to address errors and improve overall functionality.
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.