Postdoc in 3D Real-to-Sim Robot World Models and Robot Learning University of Amsterdam in Netherlands

University of Amsterdam is advertising a postdoc in 3D real-to-sim robot world models in collaboration with Toyota Core AI (Toyota Motor Europe and Toyota Research Institute). The project sits at the intersection of robot learning, world models, 3D vision, simulation, digital twins, and physical AI.

The research goal is to help robots learn safe and reliable manipulation of the physical world from only a handful of demonstrations. The post describes a pipeline combining Gaussian Splatting reconstruction with a real physics engine (Drake), real-to-sim-to-real transfer, equivariant augmentation inside the twin, and failure prediction/validation before deployment on real robots.

The application is tied to a concrete industrial use case: door-frame assembly across Toyota’s global sites. This is a 12-month position based in Amsterdam and embedded in VISLab, which is described as a large research environment with 30+ PhDs and postdocs. The post also notes the 30% ruling for internationals.

Eligibility highlights: strong 3D vision skills, robot learning experience, and the stamina to make the system work on real hardware. The post is aimed at candidates who want their research to move from paper to industrial deployment.

Deadline: 12 July 2026.

How to apply: use the University of Amsterdam vacancy page and submit the application through the recruitment portal before the deadline.