Yuval Tassa
Research Scientist at Google
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolYuval Tassa is a research scientist based in London with eight years of experience designing model-based controllers for articulated robots using optimal control, trajectory optimization, and MPC in contact-rich domains. At Google/DeepMind he has contributed to widely used open-source projects like MuJoCo and dm_control, improving performance-critical simulation kernels (e.g., CSR-based inertia computations) and hardening testing for physics-based RL environments. His background spans multibody dynamics, control theory and numerical optimization, grounded in a PhD in neuroscience and early work bridging R&D and academic research. Known for pragmatic low-level engineering alongside algorithmic research, he combines deep theoretical insight with hands-on improvements that make large simulators both faster and more robust.
8 years of coding experience
3 years of employment as a software developer
Bachelor of Science (BSc), Physics, Bachelor of Science (BSc), Physics at The Hebrew University
Hebrew, French
Stackoverflow
1reputation
0reached
1answer
0questions
Github Skills (31)
unit-testing10
simulation10
sparse-array10
c-language10
robotics10
python10
testing10
simulations10
sparse-matrix10
sparse-file10
reinforcement-learning10
mpc10
physics10
performance-optimization10
mujoco10
Programming languages (7)
Github contributions (5)
google-deepmind/mujoco_mpc
Dec 2022 - Mar 2023
Real-time behaviour synthesis with MuJoCo, using Predictive Control
Role in this project:
ML Engineer
ML EngineerContributions:78 reviews, 50 commits, 20 PRs in 3 months
Contributions summary:Yuval contributed significantly to the `mujoco_mpc` repository, primarily by adding new tasks and modifying existing ones to incorporate new features, particularly related to the Panda robot. The user also addressed code cleanup and refactoring efforts, such as removing legacy code and improving the codebase readability. They introduced the "Panda grasp-and-bring" task and also implemented additional MTS Panda+Robotiq functionality. Additionally, they made improvements to existing Quadruped tasks.
roboticsmpcpybulletsynthesismodel-predictive-control
google-deepmind/dm_control
Jan 2018 - Nov 2022
Google DeepMind's software stack for physics-based simulation and Reinforcement Learning environments, using MuJoCo.
Role in this project:Back-end Developer & Test Automation Engineer
Back-end Developer & Test Automation EngineerContributions:11 reviews, 73 commits, 3 PRs in 4 years 10 months
Contributions summary:Yuval primarily contributed to the back-end development of the `dm_control` library, specifically focusing on the MuJoCo physics simulation engine and its integration with reinforcement learning environments. They addressed critical issues related to simulation state validation, invalid states, and the handling of MuJoCo warnings within the codebase. Furthermore, the user implemented and refined testing procedures to ensure the robustness and correctness of the physics engine by introducing new tests, and improving existing tests.
simulationreinforcementsoftware-stackphysics-based-simulationsimulation-based
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.