Werner Duvaud
Co-Founder & Director Of AI at HYPR
France
Summary
🤩
Rockstar🎓
Top SchoolWerner Duvaud is a founder and AI leader with six years of hands-on experience building end-to-end agentic AI and robotics systems, currently serving as Co-Founder & Director of AI at Hypr.ai. He combines a master's in computer science with deep learning specialization and practical expertise across reinforcement learning, computer vision, robotics, electronics and MLOps, having deployed production models for autonomous drones, vehicles, and enterprise clients. Werner authored a widely used open-source MuZero implementation, contributing core trainer optimizations that improved training stability and performance. He has a track record of turning research into prototypes and funded startups—raising $5M for Hypr.ai—and often focuses on self-improving AI that “learns to learn,” bridging research rigor with product delivery.
6 years of coding experience
2 years of employment as a software developer
International exchange Machine Learning, International exchange Machine Learning at University of California, BerkeleyMaster's degree Computer science Machine Learning / Data Science, Master's degree Computer science Machine Learning / Data Science at Centrale NantesMaster's degree Management & Data Science, Master's degree Management & Data Science at Audencia
English, German, French, Spanish
Github Skills (11)
neural-network10
pytorch10
machine-learning10
deep-learning10
trainings10
modeling10
reinforcement-learning10
model-optimization9
python9
gradient-descent9
loss-functions9
Programming languages (2)
Github contributions (5)
werner-duvaud/muzero-general
Dec 2019 - May 2022
MuZero
Role in this project:
ML Engineer
ML EngineerContributions:4 reviews, 94 commits, 22 PRs in 2 years 5 months
Contributions summary:Werner refactored and improved the `trainer.py` file, which is responsible for training the neural network, indicating a focus on the core machine learning aspects of the project. The changes include modifications to the loss functions, gradient scaling, and learning rate adjustments. These modifications contribute to optimizing the neural network training and improving the overall performance of the MuZero model.
alphagodeep-reinforcement-learninggomokumodel-based-rltensorflow
werner-duvaud/openleaf-markdown-pdf
Oct 2020 - Dec 2020
Contributions:4 commits, 8 pushes, 1 branch in 2 months
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.