Research Member of Technical Staff- Post-training & Robot Learning

What you'll do

• Design RL and broader post-training pipelines, including reward design, policy optimization, supervised fine-tuning, and preference optimization.
• Develop policy evaluation and failure-analysis methods, and iterate between simulation and real-robot testing.
• Improve adaptation to new tasks with limited demonstration data and collaborate with the pre-training team on base-model capability gaps.

What they ask for

• Hands-on experience with robot systems, robotic policy learning, or autonomous systems in an industry or research setting.
• Strong understanding of robot policy learning, including imitation learning, behavior cloning, and how RL builds on them.
• Practical familiarity with robot hardware, deployment constraints, and vision and proprioception sensors.
• Hands-on PyTorch experience and ability to diagnose policy failures and distribution shift.

Nice to have

• Hands-on reinforcement learning experience applied to real or near-real environments.
• Experience with robotics or autonomous-driving industry work, teleoperation, robot middleware, simulation, or video-generation models.

Tools

PyTorch, PPO, GRPO, ROS/ROS2, MuJoCo, Isaac Sim, Genesis

AI focus

Post-train a causal video generation model on robot demonstrations to produce reliable, generalizable robot behavior; develop RL, reward modeling, preference optimization, inverse dynamics, and few-shot adaptation methods.