PhD - Scalable and Efficient Reinforcement Learning Methods for Physical AI

Join Bosch Research on the next generation of Physical AI for autonomous decision-making. You will tackle real-world constraints to create reliable, data-efficient AI agents with simulation-to-reality transfer. Collaborate with leading experts to shape scalable, deployable AI solutions for industry. Your work will bridge cutting-edge research and practical impact, aligning with Bosch’s mission to inspire and enable intelligent products and services. This PhD project offers a strong, collaborative environment and a clear path to real-world influence.

• Explore and develop new approaches for efficient autonomous decision-making in Physical AI
• Combine imitation learning, reinforcement learning, large-scale simulation, world models, and model compression
• Create AI agents across multiple embodiments with adaptation to novel situations
• Focus on resource-efficient AI with reliability and generalization for deployment on real devices
• Collaborate with academia and industry to translate research into impactful products and services

• Extensive hands-on and theoretical knowledge of reinforcement learning (RL), including Safe RL, Offline RL, and/or Multi-agent RL
• Excellent Python skills and strong experience with ML frameworks (PyTorch, Hugging Face, JAX, TensorFlow)
• Experience in software development within large teams and deploying ML-based systems; MLOps and CI/CD are a plus
• Good knowledge of C++ and ROS 2
• Publications at top ML/robotics/vision venues are advantageous

• methodical and structured work approach
• solutions-focused mindset with ownership
• proactive and initiative-taking
• reinforcement learning (RL)
• Safe RL, Offline RL, and Multi-agent RL
• Python