Robot Learning Engineer

About This Opportunity

We are representing a confidential e-commerce client in München, Germany, who is building robotics systems for large-scale fulfilment operations. They are looking for a Robot Learning Engineer to own the creation of a training corpus from human demonstration video—deciding what makes footage trainable before hundreds of hours are recorded.

This is a foundational role on a small, newly formed team. You will define the capture specification, validate ground truth under real-world conditions, build the evaluation harness that gates what enters the dataset, and run the first training experiments on manipulation policies. You will be the machine-learning voice in technical decisions before capital is deployed.

Responsibilities

• Write and enforce the capture specification: cameras, mounts, calibration, and episode-quality criteria
• Own ground-truth measurement for hand pose estimation in challenging warehouse environments (occlusion, gloves, cold)
• Build the evaluation harness that determines which recorded hours are trainable and maintain the bar as volume scales
• Post-train open robot-learning models on collected data and run task evaluations
• Reproduce published methods, validate claims against operational constraints, and steer capture strategy before spend
• Work directly with an operations lead running floor capture and a data engineer managing the pipeline

What We're Looking For

• Strong PyTorch and computer vision skills with real 3D geometry—camera calibration, pose estimation, SLAM fundamentals
• Experience debugging extrinsic calibration and understanding drift
• Ability to read research papers and reproduce results; assemble pipelines from published work
• Familiarity with current robot-learning stacks and policy classes; understanding what training demands from data
• Dataset instincts from having trained on self-collected data; awareness of failure modes that surface only at training time
• Comfortable being the sole ML decision-maker; writing down calls and revisiting them as pilot data arrives
• Proficiency using AI coding tools (Claude Code, Devin) for harness code, data plumbing, and reproduction scripts while maintaining quality
• Particularly valuable: hand-pose estimation or egocentric video in production; multi-camera rigs and time synchronization; experience post-training robot-learning models; ROS 2

How We Work with You

Candidates apply through Get A Job.ai. Our talent team will screen your application and conduct an initial interview. We then submit qualified candidates to our client for their review and interview process. Please do not contact the client directly—all communication flows through our recruiting team to protect both your candidacy and the client relationship.

Instead of a traditional cover letter, our client asks: pick a published human-video-to-robot-policy pipeline you rate, and tell us where it breaks on a warehouse floor where every hand wears a work glove.

Pay

Compensation details will be discussed during the screening process and are competitive for the München market and this level of robotics expertise.

Equal Opportunity
Get A Job.ai is committed to fair and equitable recruiting practices. We evaluate all candidates based on qualifications and role fit.