Reinforcement Learning Engineer, Grasping

3 months ago

Houston, TX, USAMid Level

Responsibilities

Train and iterate on reinforcement-learning policies for functional grasping, tool use, in-hand manipulation, and environment interaction.
Develop reward functions, curriculum strategies, and training environments in MuJoCo and Isaac Lab.
Build and refine sim-to-real transfer pipelines for physical robotic hands.
Run experiments on real robots and in simulation, debugging and evaluating policy behavior.
Evaluate and adapt state-of-the-art learning-based grasping research for deployment on a humanoid platform.
Collaborate with the software team to deploy end-to-end grasping systems.
Benchmark grasp policies across object diversity, cluttered scenes, and real-world uncertainties.
Integrate tactile sensing and feedback into robust, force-aware grasp policies.

BS, MS, or PhD in Robotics, Computer Science, Machine Learning, or a related field.
At least 2 years of hands-on reinforcement-learning experience for robotic manipulation; exceptional recent graduates may be considered.
Ability to understand and implement ideas from recent robotics and machine-learning research.
Hands-on experience training reinforcement-learning agents for manipulation, including reward shaping and policy evaluation.
Experience with sim-to-real transfer, including domain randomization, physics tuning, or real-world policy validation.
Proficiency in Python, PyTorch, JAX, and reinforcement-learning libraries such as rsl_rl or skrl.
Experience preparing meshes and collision geometries for reinforcement-learning environments in MuJoCo or Isaac Sim.
Experience with physical robotic hands, tactile sensors, contact-rich manipulation, force/torque estimation, behavior cloning, imitation learning, or diffusion-based policies is a bonus.
Publications or project work at CoRL, RSS, or ICRA on grasping or dexterous manipulation is a bonus.
Experience in a humanoid robot startup environment is a bonus.