Tianhe Yu
Palo Alto, California, United States
Summary
🤩
RockstarTianhe Yu is a Research Scientist at Google DeepMind in Palo Alto with a decade of experience building and scaling reinforcement learning systems, currently driving Gemini RL, Thinking, post‑training work and leading efforts on Gemini 2.5. He combines academic rigor as a PhD candidate in Computer Science at Stanford with practical robotics and multi‑task RL experience from collaborations at Google Brain and the Robot Learning Lab. His NeurIPS 2021 work on multi‑task offline RL reflects a strong track record in publishing, while his open‑source contributions to Metaworld show hands‑on implementation of 6DOF robotics environments, reward functions, and observation spaces. Comfortable bridging simulation and real‑world tasks, he brings both research depth and engineering execution to large RL systems.
10 years of coding experience
Github Skills (9)
gymnasium10
openai-gym10
open-ai-gym10
environmental10
robotics10
dev-environment10
environ10
python10
enviroment10
Programming languages (4)
Github contributions (5)
Farama-Foundation/Metaworld
Feb 2019 - Jan 2020
Collections of robotics environments geared towards benchmarking multi-task and meta reinforcement learning
Role in this project:
Back-end Developer
Back-end DeveloperContributions:60 commits, 5 PRs, 3 pushes in 10 months
Contributions summary:Tianhe primarily contributed by adding new environments within the repository, specifically focusing on 6DOF environments like SawyerPickAndPlace6DOFEnv, SawyerStickPull6DOFEnv, SawyerReachPushPickPlace6DOFEnv, SawyerButtonPressTopdownWall6DOFEnv. The user made changes to existing code for the new environments, including implementation of reward functions and observation spaces. The work involved modifying code related to environment setup and interaction, suggesting a focus on the functionality and behavior of the robotics environments.
manipulationmulti-taskroboticsbenchmarkingmultiagent-reinforcement-learning
tianheyu927/mil
Nov 2017 - Oct 2018
Code for "One-Shot Visual Imitation Learning via Meta-Learning"
Contributions:24 commits, 3 PRs, 18 pushes in 11 months
pytorchmeta-learningone-shotrobustnessgeneralization
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.