SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Rohlik is Central Europe's leading e-grocer, operating across five countries with over a million customers. The company is building autonomous fulfillment capabilities through AI, robotics, and computer vision. This role is part of a newly created team focused on capturing and labeling human manipulation data to train robot learning policies for warehouse operations.
You will own the complete pipeline for building a production-grade robot learning dataset:
**Capture Specification**: Design and enforce the camera setup, mounts, calibration standards, and quality gates that determine which warehouse footage is trainable. You write the spec before hardware procurement and validate it against floor reality.
**Ground Truth Annotation**: Hand pose estimation is the critical signal for transferring human demonstrations to robot policies. You will own the measurement and labeling strategy, accounting for real-world challenges like work gloves, occlusion, and cold warehouse environments where published models typically struggle.
**Evaluation Harness**: Build the system that determines which hours of footage meet the training bar. As volume scales, you maintain data quality standards and decide what passes into the training corpus.
**First Training Runs**: Post-train open-source robot learning models on the collected dataset and run initial task evaluations to validate the corpus quality.
**Technical Leadership**: Stay current with the robot learning field, reproduce published claims, and steer data capture strategy based on emerging research before capital is committed.
You will work on a small team: an operations lead managing floor capture, a data engineer owning the pipeline, and you as the ML voice. This is not a research role—the focus is making data trainable and training the first policies, not publishing papers.
The company uses AI agents (Claude Code, Devin) as standard tools; you will direct them for implementation work while maintaining technical quality.
**Requirements:**
- Strong PyTorch and computer vision expertise with real 3D geometry (camera calibration, pose estimation, SLAM fundamentals). Hands-on experience debugging extrinsic calibration drift.
- Ability to read academic papers and reproduce claims in production settings.
- Deep familiarity with current robot learning stacks and policy classes; understanding of what training pipelines demand from data.
- Prior experience training on self-collected datasets and knowledge of real-world data failure modes.
- Comfortable as the sole ML decision-maker; ability to document decisions and iterate based on pilot results.
- Particularly valued: hand-pose estimation or egocentric video in production; multi-camera rig synchronization; post-training robot learning models; ROS 2 experience.
**90-Day Success Metrics**: Capture spec written and first rigs deployed; pilot evaluation gates defined with quantitative targets; initial footage passes harness; data-driven recommendations for scaling to hundreds of hours.
