Postdoc on 3D Real-to-Sim Robot World Models with Toyota
- Deadline
- 4 Dec 2026 · 12 days left
- Location
- Amsterdam, Netherlands
- Type
- Postdoc
- Compensation
- EUR 3,546 – 5,538 per month (Scale 10 (UFO Researcher 4), CAO Dutch Universities)
- Duration
- 12 months · 1 year
- Working language
- English
- Posted
- 19 Nov 2026
- Verified
- 19 Nov 2026
About this opportunity
The University of Amsterdam's Faculty of Science is hiring a postdoctoral researcher to build physics-grounded robot "digital twins". You would join Efstratios Gavves's group and the VISLab, working closely with the Toyota Core AI team. The project combines Gaussian Splatting 3D reconstruction with physics engines such as Drake so that robots can learn manipulation skills from only a few real demonstrations. The practical target is automating door-frame assembly in Toyota factories, together with Toyota Motor Europe and Toyota Research Institute.
What is offered
- A 12-month temporary contract at 38 hours per week, starting as soon as possible.
- Salary on scale 10 (UFO profile Researcher 4): €3,546 to €5,538 gross per month, depending on experience.
- An 8% holiday allowance and an 8.3% year-end allowance on top of salary, under the Collective Labour Agreement of Dutch Universities.
- The 30% ruling, a Dutch tax benefit, may apply to non-Dutch applicants.
- Direct collaboration with industry, the chance to publish at top venues (CVPR, ICLR, NeurIPS, ICML, ECCV), contribute to joint patents, and co-supervise PhD students.
Main tasks:
- Build object-centric 3D world models that combine Gaussian Splatting with physics simulators.
- Develop real-to-sim-to-real pipelines that create simulation environments from video or images.
- Develop scene completion and occlusion-handling methods for objects that are only partly visible.
- Design equivariant data augmentation for imitation learning.
- Build real-time failure-prediction and monitoring modules based on the digital twin.
Who can apply
- A PhD in computer vision, machine learning, robotics or a related field, with a strong publication record at top venues.
- Expertise in 3D reconstruction, differentiable rendering or Gaussian Splatting.
- Experience with robot learning (imitation or reinforcement learning) in simulation or on real hardware.
- Proficiency in PyTorch or JAX and in a physics simulator such as Drake, MuJoCo or PyBullet.
- A proven ability to lead research independently and to work across institutions.
- Professional-level English.
- Experience with vision foundation models or generative 3D reconstruction is a plus.
How to apply
- Apply online with the "Apply" button on the vacancy page by 4 December 2026.
- Send one PDF that contains:
- a detailed CV (up to 2 pages), with months for education and work dates, your top publications, and a motivation section (up to 1 page);
- a research statement (up to 2 pages);
- the names and email addresses of two referees.
- The selection process may include a knowledge security check.
- For questions, contact Efstratios Gavves (Associate Professor) using the contact link on the vacancy page.
Requirements
-
Open to all nationalities
-
Early-career researcher
-
Minimum degree: PhD
Postdoc on 3D Real-to-Sim Robot World Models with Toyota
4 Dec 2026 · 12 days left