The Opportunity
Want to work on the machine learning models that actually control robots? We're looking for an engineering student to join us on post-training our robot foundation models, starting with a 3-month internship and continuing as a working student (minimum 1 year total).
This is a hands-on role at the core of our ML work. Together with the team, you'll help finetune large foundation models for robotics and improve the real-world performance of our control policies. You'll get to read recent papers, implement them in code, and apply them on real robots to see the results, and along the way you'll learn how large robot foundation models are trained, deployed, and improved. For the right person, this can also open up thesis topics down the line.
Your Responsibilities
-
Help optimize our fine tuning pipelines and improve the performance of our control policies.
-
Read recent papers, help us implement them in code, and test them on real robots.
-
Experiment with new post-training techniques for our models.
-
Contribute to and improve our existing evaluation and benchmarking harnesses: compute metrics, compare model versions, and report what you find.
-
Improve the tooling and workflows that make training and fine tuning jobs easier to start, configure, and debug.
-
Automate recurring post-training work, such as hyperparameter and ablation sweeps.
Requirements
-
Currently enrolled in a Master's programme in Robotics, Computer Science, Machine Learning, Electrical Engineering, or a related field (early in your studies is welcome).
-
Exposure to robotics or control, for example through lectures:control theory, state estimation, ROS/ROS 2, coordinate frames and transforms, or sensor data.
-
Good software fundamentals and solid Python (numpy, pandas, torch), comfortable working in Linux.
-
Some hands-on ML experience, preferably deep learning.
-
A hands-on, proactive mentality, happy to dig into messy data and figure things out.
-
Available for a 3-month internship followed by a working-student role, around 1 year minimum in total.
Nice to Have
-
Any exposure to robot foundation models (WAMs, VLAs VLMs, LLMs), fine tuning large models, post-training methods (LoRA/PEFT), PyTorch/multi-GPU training, or robotics datasets like LeRobot.
-
Genuine interest in data and statistics.
Benefits
-
Work with a world-class team in a flat hierarchy, with direct collaboration alongside the founders and engineering team
-
Opportunity to make a real impact by working on cutting-edge robotics and AI systems
-
Fast growth potential in a rapidly evolving company and industry
-
International office environment with English as the official working language
Recruiting Process
Your recruiting partner for this role is Madhulika (she/her). You can expect a screening interview and upto 2 rounds of interviews including an onsite visit to our office in Munich.
We hire across backgrounds, identities, and experiences, and we are committed to a workplace where everyone belongs. Discrimination has no place here.
If you need any accommodations during the recruiting process, just reach out to your recruiting partner.
