About the job
As a research engineer in the semantic understanding and reasoning group (CR/AIR4) at Bosch Corporate Research, you will develop next-generation methods for training large behavior models for intelligent cyber-physical systems. Your work will focus on how large-scale AI models can acquire robust, generalizable, and goal-directed behaviors through reinforcement learning, multimodal experience, and interaction with learned or simulated environments.
Responsibilities
Investigate how predictive models of environment dynamics, latent state, and agent-environment interaction can support policy learning, planning, behavior synthesis, and evaluation.
Leverage world-model-based rollouts for scalable training.
Use imagined trajectories for efficient policy improvement.
Develop validation frameworks that assess generalization, robustness, and safety before real-world deployment.
Contribute to the design of architectures that connect representation learning, latent dynamics modeling, reinforcement learning, and large-scale behavior modeling.
Build infrastructure needed for pretraining, simulation-based learning, fine-tuning, and benchmarking in Bosch-relevant environments.
Collaborate closely with AI researchers, robotics experts, control engineers, and domain specialists.
Qualifications
Minimum
Excellent MSc in Computer Science, Machine Learning, Robotics, Control, or related technical fields
Expertise in reinforcement learning and sequential decision-making for complex environments
Experience with model-based, offline, hierarchical, imitation, or constrained RL
Training large-scale behavior or policy models from multimodal data and interaction
Designing methods for long-horizon optimization, generalization, and robust adaptation
Solid understanding of world models, latent dynamics, and sequence or generative models
Using predictive models for imagination-based training, rollouts, and planning
Experience with latent-state modeling, uncertainty-aware prediction, and validation
Experience with large-scale deep learning and transformer-based or multimodal models
Representation learning across visual, temporal, action, language, or sensor modalities
Strong Python skills and experience with PyTorch, TensorFlow, or JAX
Experience with simulation platforms (e.g., Isaac Sim, CARLA, MuJoCo, Habitat)
Familiarity with distributed training, benchmarking, and reproducible pipelines
Experience with Docker, Git, CI/CD, and multi-GPU or cloud infrastructure
Preferred
PhD in Machine Learning, Reinforcement Learning, Robotics, Generative AI, or related areas preferred
Strong publication record in leading AI, machine learning, and robotics venues such as NeurIPS, ICLR, ICML, CoRL, RSS, ICRA, AAAI, IJCAI, or similar
Strong scientific mindset with a proven publication record in top-tier AI and robotics venues
Ability to translate cutting-edge research into practical, value-creating innovations
Interest in behavior validation, robustness testing, sim-to-real, and safety
Interest in connecting data-driven learning with physically grounded reasoning
Interest in large behavior models as transferable, reusable AI components
Linking large-model training with policy learning and environment interaction
German language skills