Robot Learning Engineer

8 days ago

Prague, CzechiaSenior

Responsibilities

Define and enforce camera, mounting, calibration, recording, and episode-quality specifications for warehouse data capture.
Establish hand-pose ground truth and address occlusion, gloves, and other warehouse-specific measurement challenges.
Build an evaluation harness that determines which recorded hours are suitable for robot-policy training.
Post-train open robot-learning models on the collected data and run initial task evaluations.
Read and reproduce published research, validate its claims in warehouse conditions, and guide future data-capture investment.
Serve as the primary machine-learning voice on a small team and make data-informed decisions about pilot and corpus quality.

Hands-on experience with PyTorch and computer vision involving real 3D geometry.
Knowledge of camera calibration, pose estimation, and SLAM fundamentals, including debugging extrinsic calibration.
Ability to read and reproduce published research and assess whether its claims hold in practical warehouse conditions.
Familiarity with robot-learning policy classes and the data requirements of current robot-learning systems.
Experience training on self-collected data and recognizing dataset failure modes.
Particularly relevant experience includes hand-pose estimation or production egocentric video, multi-camera rigs and time synchronization, post-training robot-learning models, or ROS 2.
Ability to work independently as the sole machine-learning specialist and direct AI coding agents while maintaining quality standards.

Ownership of the physical-AI capture specification and first training runs from day one.
Opportunity to build a large-scale robot-learning data asset in grocery fulfillment and manipulation.
Off-the-shelf hardware and a managed cloud estate are already available.
The role works directly with warehouse operations and a newly created small team; specific additional benefits are provided via the employer's benefits page.

501-1,000 employees