Robot Machine Learning Engineer (Werkstudent*in) – Humanoid Robotics Startup

As the technological capabilities of humanoid (and other generalist) robots advance, there is an increasing demand among German companies to integrate these robots into their workforce. This is what RoboService does: We provide “Robots-as-a-Service” and provide companies with the robots that they need to automate their tasks. As an ML working student, you’ll collect teleoperation data, train policies, and deploy them onto real hardware—quickly. You’ll work directly with our founders and ship to the lab every week.

What you’ll do:

Below are some of the core focus areas for Machine Learning Engineers at RoboService:

1. Collect & curate teleoperation datasets (RGB/Depth + proprioception + force/torque): define schemas, ensure time sync (PTP/NTP), and set quality gates.
2. Train and evaluate imitation-learning / diffusion policies (PyTorch) for manipulation; stand up baselines and compare against strong references.
3. Build fast simulation loops (Isaac Sim/Gym or Gazebo) for augmentation and rapid iteration; apply domain randomization for Sim2Real.
4. Optimize inference on robot (TorchScript/ONNX/TensorRT) with latency/throughput targets on Jetson/x86.
5. Integrate policies into ROS 2 nodes; define safe fallbacks, gating, and E-stop pathways with controls teammates.
6. Explore Vision-Language-Action approaches for task generalization and evaluation.