> Markdown version of [/jobs/ext/4496937-robot-learning-engineer-human-motion-data](https://www.wearedevelopers.com/jobs/ext/4496937-robot-learning-engineer-human-motion-data). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Robot Learning Engineer, Human Motion Data

- **Company:** ease GmbH
- **Location:** München, Germany
- **Salary:** €62,500.0 - €79,000.0
- **Contract:** Permanent contract
- **Skills:** C++ (Programming Language), Python (Programming Language), Kinematics, Reinforcement Learning, Pytorch, Transfer Learning, Stable Diffusion
- **Published:** November 1, 2026
- **Apply:** https://www.xing.com/jobs/garching-robot-learning-engineer-human-motion-data-158868845

## About the Role

* Master's or PhD in robotics, CS, ML or similar, with hands-on experience from industry or research
 * Imitation or reinforcement learning on real robots (e.g. Diffusion Policy, ACT, PPO)
 * Strong Python and PyTorch; C++ is a plus
 * Physics simulation and sim-to-real transfer
 * Solid kinematics, dynamics and control, EASE is an equal opportunity employer. Applicants need a valid work permit for the EU.

## Description

Our exoskeletons are deployed in real warehouses and production plants and capture what video-only datasets miss: body pose, joint kinematics, interaction torques, vision, and hand forces from real human work. Your job is to turn that data into skills our humanoid can actually perform, so that human workers and robots can work side by side on the same tasks.

What you'll do

 * Map human motion onto humanoid kinematics and close the embodiment gap
 * Train manipulation and whole-body policies with imitation and reinforcement learning
 * Evaluate in Isaac Lab or MuJoCo and on our in-house humanoid, and measure the value of our data
 * Define what data training needs and help build our training setup, including GPU compute
 * Turn new research on robot foundation models and VLAs into prototypes