Senior AI Research Engineer (m/w/d) Foundation Models

About the role:
The AI Research Division of Agile Robots is looking for a Senior AI Research Engineer focused on learning-based robot policies and multimodal foundation models for embodied intelligence. The role centers on designing representations and objectives that map perception and language to robot behavior under real-world uncertainty, with ownership of model-level decisions and behavioral outcomes.

Your Responsibilities:
• Policy Design: Design and train vision- and language-conditioned robot policies using imitation learning, diffusion-based approaches, or transformer architectures, with clear linkage between representation, objective, and resulting behavior.
• Model Architecture: Define and adapt model architectures, inductive biases, and training objectives for embodied learning under uncertainty beyond standard fine-tuning workflows.
• Behaviour Analysis: Analyze policy behavior and failure modes under distribution shift, embodiment constraints, and real-world noise, and refine models to improve robustness and generalization.
• Learning Pipelines: Own end-to-end learning pipelines from data assumptions through training, evaluation, and controlled real-world validation in collaboration with robotics teams.
• Integration Collaboration: Collaborate closely with robotics, perception, and platform teams to ensure learned policies align with action spaces, timing constraints, and system-level requirements.
• Research Awareness: Track and critically evaluate advances in robot learning, multimodal foundation models, and vision-language-action systems, assessing their applicability to embodied intelligence.

Essential Skills:
• Education: Master’s or PhD in Computer Science, Robotics, AI, or a related technical field.
• Robot Learning: Hands-on experience designing and training learning-based policies for robotics using imitation learning, reinforcement learning, or hybrid approaches.
• Multimodal Models: Experience adapting or designing transformer-based, diffusion-based, or vision-language-action models for embodied tasks.
• Modelling Depth: Ability to reason about representations, objectives, inductive biases, and behavioral trade-offs, including explaining why a model is structured in a particular way.
• Behavioural Ownership: Demonstrated responsibility for model-level outcomes, including debugging policy failures and improving robustness in real-world or semi-real environments.
• ML Engineering: Strong Python skills and hands-on experience with modern ML frameworks such as PyTorch or TensorFlow, including implementation of training loops and experimentation workflows.
• Robotics Context: Understanding of how learned policies interact with perception outputs, action spaces, and physical constraints in robotic systems.

Beneficial Skills:
• Large-Scale Training: Experience with distributed training, GPU clusters, or large-scale experimentation workflows.
• Predictive Components: Experience leveraging predictive or world-model elements to improve policy robustness or long-horizon behavior.
• Embodied Evaluation: Experience validating learned policies on real robotic systems or in high-fidelity simulation with sim-to-real considerations.
• Research Output: Publications, patents, or deployed systems in robotics, multimodal learning, or embodied AI.

What we offer:
• Dynamic high-tech company combined with financial soundness and world class investors.
• Join an interdisciplinary, international team with 60+ different nationalities in a collaborative work environment.
• Lots of development opportunities in the context of our continued growth.
• Challenging tasks and impactful projects alongside experts that enable professional and personal growth.
• Corporate Benefits Program that covers health, mobility and learning with 100 € net per month.
• Modern office facilites with a rooftop terrace overlooking Munich, free drinks & fruits, and regular company events contribute to a good working environment.