Juan Rocamonde
Co-Founder & CEO at Radiela
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolJuan Rocamonde is a founder and research-driven CEO based in San Francisco with a decade of experience at the intersection of machine learning, AI safety, and applied research. He builds steerable and explainable AI systems as a Research Fellow at FAR AI and directs strategy on transformative-AI implications at Newton Turing, translating rigorous research into practical governance and tooling. His background spans theoretical physics and advanced mathematics (Cambridge, UCL), with hands-on ML engineering contributions to notable open-source RL projects like stable-baselines3 and imitation—improving type safety, vectorized-environment behavior, and Apple Silicon support. Juan has applied ML in diverse domains from antibody property prediction at AstraZeneca to falsifying BSM physics at CERN, reflecting a talent for turning deep technical ideas into reproducible code and experiments.
10 years of coding experience
1 year of employment as a software developer
Part III of the Mathematical Tripos, Part III of the Mathematical Tripos at University of CambridgeUniversity College London
Spanish, Galician, Portuguese, English
Github Skills (18)
imitation-learning10
pytorch10
python10
machine-learning10
sac10
reinforcement-learning10
typehinting10
dqn10
tdd10
type-checking10
gymnasium9
logging8
pytest8
testing8
documentation7
Programming languages (6)
Github contributions (5)
HumanCompatibleAI/imitation
Jul 2022 - Jan 2023
Clean PyTorch implementations of imitation and reward learning algorithms
Role in this project:
ML Engineer
ML EngineerContributions:215 reviews, 95 commits, 35 PRs in 6 months
Contributions summary:Juan primarily contributes to the implementation and refinement of imitation and reward learning algorithms within the `imitation` repository. Their work involves modifying existing code and adding support for new features, such as Apple Silicon installation and warm-starting using pre-trained policies. A key aspect of their contribution is addressing type-related issues and maintaining code quality through consistent type hinting and code style enforcement. The user's commits also improve the codebase's organization, including documentation and test suite enhancements.
pytorchimplementationsreinforcement-learningcleanmachine-learning
DLR-RM/stable-baselines3
Sep 2022 - Nov 2022
PyTorch version of Stable Baselines, reliable implementations of reinforcement learning algorithms.
Role in this project:ML Engineer
ML EngineerContributions:8 reviews, 6 commits, 6 PRs in 2 months
Contributions summary:Juan primarily contributed to improving the type annotations and fixing return types within the library's codebase, specifically concerning algorithms like DQN, SAC, and TD3. They corrected type annotations for the `replay_buffer_class` argument and the return types of `load` and `learn` methods in the base algorithm. Additionally, they addressed a duplicate key error in the `HumanOutputFormat` logger and the return type of `evaluate_actions` in `ActorCriticPolicy`. Furthermore, the user's commits ensured correct behavior when using vectorized environments, which reflects a focus on maintaining a high-quality RL library.
pythonstable-baselinesrobustnessgsdesde
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.