Ryan Pégoud
Research Engineer — RL, Knowledge Distillation, GPU kernels
Montpellier, France · ryan.pegoud@outlook.com · GitHub · LinkedIn
Open to research engineer roles in London.
Download CV ↓About
I'm a research engineer working on inference efficiency and model compression, with a background in reinforcement learning. I completed an MSc in Computational Statistics and Machine Learning at UCL (Distinction), where my thesis "neural memory augmentation for in-context learning in agentic language models" was supervised by Tim Rocktäschel. I also hold an MEng in computer engineering.
Recently, I was a Visiting Researcher at BMW Group, where my work focused on white-box knowledge distillation of reasoning VLAs for autonomous driving.
I write about GPU kernels and reinforcement learning on this site and on Towards Data Science. My articles on Triton, JAX, and transformers have been read over 100,000 times and picked up multiple Editor's Picks.
Publications
Conference Papers
Ryan Sullivan, Ryan Pégoud, Ameen Ur Rehman, Xinchen Yang, Junyun Huang, Aayush Verma, Nistha Mitra, John P. Dickerson. "Syllabus: Portable Curricula for Reinforcement Learning Agents". Reinforcement Learning Conference (RLC), 2025.Outstanding Paper Award — Tooling, Environments, and Evaluation for RL.
Workshop Papers
Ryan Pégoud, Thibault Lahire. "Better Gradient Steps for Deep On-Policy Reinforcement Learning". Aligning Reinforcement Learning Experimentalists and Theorists Workshop, ICML, 2024.
Writing
12 technical articles on Towards Data Science and Data Science Collective, covering RL, NLP, JAX, Transformers, Triton, and GPU programming — 100k+ reads and multiple Editor's Picks. A selection is featured on this site.