Develop and improve perception models for autonomous mobile robots, including object detection, obstacle classification, and free-space estimation.
Drive data-centric machine learning by scaling automated labeling, leveraging fleet data, and closing the sim2real gap through synthetic data and simulation-based evaluation.
Own model performance by defining benchmarks, tracking key metrics, and continuously improving real-world and benchmark results.
Deploy and validate perception models within a ROS-based robot software stack, ensuring reliable performance on real vehicles.
Collaborate closely with navigation, controls, and fleet software teams to deliver robust perception capabilities.
Requirements
2–5 years of professional and hands-on experience in robotics perception, computer vision, or machine learning.
Strong Python skills; experience with C++ is a plus.
Hands-on experience training and evaluating neural networks (PyTorch or similar) with a data-centric approach.
Practical experience with synthetic data, simulation for ML, auto-labeling, active learning, or domain adaptation.
Familiarity with ROS and Linux-based development environments.
What We Have to Offer
Cutting-edge robotics technology navigating and manipulating without rails and guides.
An international team with more than 25 nationalities.
A loft-style office near the S-Bahn station "Hirschgarten" with a workshop and testing area, and a free daily cooked lunch.
Hybrid working model, attractive holiday, Wellpass, health care, and compensation packages.