Vincent Moens
Member Of Technical Staff And Founding Team at Periodic Labs
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolVincent Moens is an ML research scientist and software engineer with eight years of experience building production-grade reinforcement learning infrastructure and contributing to core PyTorch projects while now serving on the founding technical team at Periodic Labs in London. He holds an M.D. and a Ph.D. in Computational and Cognitive Neuroscience, blending clinical insight with deep expertise in generative models, time-series transformers, AutoML/meta-learning and (Bayesian) RL. At Meta he maintained TorchRL and made substantive backend, testing and memmap contributions across high-profile PyTorch repos (pytorch/pytorch, vision, tensordict and tutorials), fixing core bugs and improving test reliability. Comfortable across Julia (preferred), Python, Matlab, R and Stan, he pairs rigorous algorithmic research—change-detection and sequential decision making—with hands-on engineering for scalable, GPU-accelerated systems. He also advises on reliable RL control for space systems, a hint of how his work spans from theoretical models to safety-critical applied domains.
8 years of coding experience
3 years of employment as a software developer
Doctor of Philosophy - PhD Computational neurosience, Doctor of Philosophy - PhD Computational neurosience at Université catholique de Louvain
English, French, Italian
Github Skills (26)
pytorch10
debug10
pytest10
python10
testing10
gymnasium10
openai-gym10
machine-learning10
reinforcement-learning10
tensorflow10
neural-network10
computer-vision10
tensor10
maintenance9
packaging9
Programming languages (9)
Github contributions (5)
pytorch/tensordict
Oct 2022 - Mar 2023
TensorDict is a pytorch dedicated tensor container.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:20 releases, 509 reviews, 372 commits in 5 months
Contributions summary:Vincent contributed to the foundational code for TensorDict, a PyTorch-based tensor container. Their initial commit introduced the core logic of the memmap feature, including file handling and data persistence. The user subsequently added code to the project, including a file called setup.py, and refactored the project's testing, packaging and build scripts. The user also implemented testing related to Memmap and fixed various features related to the core functionality.
pytorchdeep-learningdedicateddockertensor
pytorch/rl
Mar 2022 - Jan 2023
A modular, primitive-first, python-first PyTorch library for Reinforcement Learning.
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:23 releases, 1926 reviews, 807 commits in 10 months
Contributions summary:Vincent's contributions focused on TorchOpt compatibility fixes, addressing potential issues with the `torchrl/envs/vec_env.py` module. These fixes involved modifications to ensure compatibility with TorchOpt, a library used for optimization in PyTorch. The code changes also include the renames of 'tensor_dict' to 'tensordict', which is a common practice in code maintenance to improve readability. Furthermore, the user implemented replay buffer functionality for storing trajectories, contributing to the reinforcement learning setup within the project.
pytorchpythonreinforcement-learningprimitivedeep-reinforcement-learning
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.