Aymeric Roucher
Co-Founder at Odile Jacob
Paris, Ile-de-France
Summary
🤩
Rockstar🎓
Top SchoolAymeric Roucher is a machine learning engineer and founder with five years of applied ML experience, based in Paris. He led Hugging Face’s Agents effort—launching the smolagents library (now 20k+ stars) and building evaluation suites like ScreenSuite and Screenenv—while improving agent tooling and Python execution for ReAct-style agents. Previously he engineered production ML and geospatial risk models at Descartes Underwriting, shipping Gen-AI prototypes and a frost insurance product that generated revenue. His background spans robotics state estimation for assistive exoskeletons and disciplined leadership from military special-forces training, blending rigorous engineering with operational grit. He holds degrees from École Polytechnique and Cambridge and co-founded Verso AI while continuing hands-on open-source development.
6 years of coding experience
4 years of employment as a software developer
Baccalauréat : Highest honors, Baccalauréat : Highest honors at Lycée Henri IVBSc + MSc Computer Science | X17, BSc + MSc Computer Science | X17 at École PolytechniqueMPhil, MPhil at University of CambridgeBSc Mathematics & Physics | MP*, BSc Mathematics & Physics | MP* at Lycée Sainte-Geneviève
English, German, French, Chinese
Github Skills (13)
transformers10
debugging10
pytorch10
machine-learning10
develop10
agent10
nlp10
python10
natural-language-processing10
deep-learning9
open-source9
language-model9
testing8
Programming languages (9)
Github contributions (5)
huggingface/smolagents
Dec 2024 - Mar 2026
🤗 smolagents: a barebones library for agents that think in code.
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:7 releases, 670 reviews, 375 PRs in 1 year 2 months
Contributions summary:Aymeric appears to have focused on the development and refinement of a barebones agent library for thinking in Python code. They implemented the initial repository structure and console output improvements for better debugging and readability. Furthermore, the user made changes for the management and execution of agents and tools within the framework.
huggingface/transformers
Feb 2024 - Apr 2025
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
Role in this project:ML Engineer
ML EngineerContributions:90 reviews, 73 PRs, 532 pushes in 1 year 1 month
Contributions summary:Aymeric's primary contribution involves significant modifications to the `transformers/agents` module, which focuses on building and utilizing agents within the Hugging Face ecosystem. These changes include the implementation of streaming capabilities for ReAct agents, allowing additional imports in code agents, and enhancements to the Python interpreter to support classes, exceptions, and other improvements. The user has also improved planning for the agents and added new tools, indicating a focus on expanding the capabilities and usability of the agent framework. They've also made some small fixes and added tests.
audioinferencemachine-learning-modelsmultimodaltransformers
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.