Robot Learning Engineer: Teleoperation & Policy

Lightspeed; Northbrook (IL); Vor Ort; USD 100.000 - 120.000; Vollzeit

Zusammenfassung

LightSpeed Build Technologies is seeking a Robot Learning Engineer to teach our robots complex manipulation tasks, from demonstration to deployed behavior. You will own the path from demonstration to deployed behavior, building teleoperation systems, turning data into trained policies, and proving those policies on real robots with dexterous hands and custom tooling.

You will design data collection protocols, retarget human motion, and shape data strategies while guiding technicians who run

Qualifikationen

• 3+ years hands-on robot learning or production ML experience.
• Experience training policies from demonstrations on physical robots.
• Experience collecting teleoperation or demonstration data.
• Strong PyTorch skills for end-to-end training, evaluation, debugging.

Aufgaben

• Build and improve teleoperation pipelines using leader arms and VR hand tracking.
• Retarget human hand motion to a dexterous robot hand, in simulation then on hardware.
• Design data collection protocols for consistent, trainable demonstrations.
• Define data worth and guide technicians during collection sessions.
• Train and fine-tune manipulation policies using imitation learning and VLA models.
• Choose policy architectures for multi-step, contact-rich tasks.
• Combine teleoperation data with video to improve policies.
• Iterate from data to trained policy to real-robot trial.
• Define evaluation protocols and measure task success on real robots.
• Diagnose failures across data, model, and hardware; fix root causes.

Kenntnisse

Robot learning

Imitation learning

PyTorch

Python

C++

ROS/ROS2

Linux

Git

Ausbildung

MS or PhD in Robotics or ML

Tools

MuJoCo

Isaac Sim

Docker

rosbag / MCAP

Jobbeschreibung

LightSpeed Build Technologies is seeking a Robot Learning Engineer to teach our robots complex manipulation tasks, from demonstration to deployed behavior. You will own the path from demonstration to deployed behavior, building teleoperation systems, turning data into trained policies, and proving those policies on real robots with dexterous hands and custom tooling.

You will design data collection protocols, retarget human motion, and shape data strategies while guiding technicians who run