Robot Learning Engineer

About the role

Rohlik is building the autonomous grocery, and needs a Robot Learning Engineer to create a trainable video corpus from warehouse operations. You will define how footage is captured, labelled and turned into manipulation policies for robots that move fresh food from producers to households.

Key responsibilities

• Define and enforce the capture specification – camera setup, mounts, calibration and episode disqualification criteria.
• Design the ground‑truth pipeline for hand‑pose estimation in a cold, glove‑filled warehouse environment.
• Build the evaluation harness that decides which recorded hours are trainable and maintain the trainability bar as data volume grows.
• Run the first training cycles on the collected data and evaluate robot‑learning policies.

Required profile

• Hands‑on experience with PyTorch and computer‑vision systems that operate on real 3‑D geometry.
• Practical knowledge of camera calibration, pose estimation and SLAM fundamentals, including fixing extrinsic drift.
• Ability to read research papers, reproduce results and judge their applicability to a chilled warehouse setting.
• Experience collecting your own data and understanding failure modes that appear only during training.

Required skills

• PyTorch
• Computer vision with 3D geometry
• Camera calibration & pose estimation
• SLAM fundamentals
• Hand‑pose estimation or egocentric video pipelines
• Multi‑camera rigs and time‑synchronisation
• ROS 2

What we offer

• Early ownership of the capture spec and the first robot‑learning training runs.
• Access to a unique data asset in fresh‑grocery manipulation, a domain few teams worldwide tackle.
• Off‑the‑shelf hardware, managed cloud infrastructure and a mandate that is already funded.
• Opportunity to influence capital decisions directly from technical judgement.