Robotics Intern, Embodied Spatial Reasoning

This role is based in Zurich, Switzerland.

What you'll be doing:

• Work closely with robotics researchers and engineers to define a project focused on embodied spatial reasoning.
• Design, implement, and evaluate algorithms for geometry-aware Action Chain-of-Thought (Action CoT) representations.
• Develop controlled spatial robustness benchmarks to quantify how Action CoT affects generalization and robustness in real and simulated environments.
• Apply your methods to improve planning for simulated or real robotic systems.
• Integrate your algorithms into NVIDIA’s robotics ecosystem.

What we need to see:

• Experience in Embodied AI, VLA/VLM systems, diffusion models, or World Foundation Models (WFM).
• Currently enrolled in a PhD, Master’s degree, or equivalent in Computer Science, Robotics, Electrical Engineering, or a related field.
• Strong programming skills in Python and familiarity with deep learning frameworks (PyTorch preferred).
• Demonstrated research or internship experience, ideally with publications in top-tier conferences.
• Excellent communication and collaboration skills.

Ways to stand out from the crowd:

• Experience deploying LLMs, VLA/VLM systems, diffusion models, or WFMs in real-world or production-scale systems.
• Hands-on experience with NVIDIA Cosmos, Isaac Sim, or Isaac Lab for synthetic data generation, simulation, and physical AI development.