Applied Robot Learning Engineer / RL Specialist
$153k - $234k • Mountain View, California, United States of America • Full-time
Posted 10d ago
Job Location
Mountain View, California, United States of America
Tech Stack
Remote Work Policy
On-site
Employment Type
Full-time
Categories
AI Engineer
About the job
We are seeking an experienced applied robotics engineer with a strong background in reinforcement learning (RL) to deploy learning-based policies for complex manufacturing tasks. This role involves enhancing world models, tackling sim-to-real challenges, and developing robust policies for manipulation, failure recovery, and optimal execution. You will work hands-on with robotic hardware, requiring flexibility to support broader engineering and experimental tasks as project priorities shift. The ideal candidate blends RL fundamentals with practical robotics engineering and some experience in bimanual manipulation. Approximately 80% of your time will be dedicated to applied engineering, policy implementation, robot experiments, evaluation, and deployment, with the remaining 20% focused on research and technical exploration, emphasizing measurable improvements on physical robots.
Responsibilities
- Develop, fine-tune, and evaluate VLA models and learning-based policies on physical robotic systems.
- Apply reinforcement learning to enhance manipulation success rates, execution efficiency, robustness, and failure recovery.
- Implement and evaluate RL approaches such as PPO, SAC, residual RL, policy adaptation, or learned recovery policies.
- Design reward functions and success detectors, including learned or VLM-based reward models.
- Apply sample-efficient methods for real-world training, including RL from demonstrations, offline-to-online RL, and human-in-the-loop corrections.
- Set up safe real-robot RL, incorporating action limits, workspace constraints, automated resets, and supervised exploration.
- Diagnose sim-to-real gaps and improve real-world performance using techniques like domain randomization, system identification, and real-world fine-tuning.
- Apply learned policies to manipulation tasks, including bimanual setups and coordinated robot execution.
- Design evaluations, analyze failure modes, and measure improvements in success rate, cycle time, generalization, and repeatability.
- Work directly with robotic hardware to run experiments, debug policies, and validate improvements.
- Supervise data collection, identify data-quality issues, and coordinate corrections to demonstrations or datasets.
- Collaborate with the simulation team on benchmarking, model evaluation, and other engineering tasks.
Requirements
- Ph.D. in Computer Science, Robotics, or a related discipline, or a Master’s degree with 3+ years of industry experience.
- Strong applied expertise in reinforcement learning for robotics, including policy optimization, reward design, training, and evaluation.
- Proven experience deploying or validating learned policies on physical robots; simulation-only experience is insufficient.
- Experience with sim-to-real transfer and diagnosing discrepancies between simulated and real-world execution.
- Proficiency with Git, Python, PyTorch, and ROS/ROS 2.
- Hands-on familiarity with edge compute platforms (e.g., NVIDIA Jetson) and robotics MLOps ecosystems (experiment tracking, dataset versioning, visualization tools like Foxglove).
- Experience with sensor calibration (eye-in-hand/eye-to-hand camera calibration), force-torque sensors, or teleoperation data collection setups.
- Understanding of imitation learning, robot kinematics, control interfaces, and manipulation workflows.
- Comfort working directly with robots and taking ownership of experiments, model training, evaluations, data quality, debugging, and deployment.
- Adaptability and willingness to contribute beyond core model development to meet project goals.
