Postdoctoral Fellow / Research Scientist — Embodied AI & Spatial AI & World Models

We are hiring across several tightly coupled themes that map onto the most urgent open problems in embodied AI:

1. Large-Scale Multimodal & World Models. Design latent autoregressive or diffusion simulators that learn from web-scale vision-language data and closed-loop robot logs.
2. Vision-Language-Action (VLA) & Vision-Language-Navigation (VLN). Create transformer backbones that plan, ground and execute long-horizon tasks in natural language.
3. Reinforcement Learning (RL) at Scale. Push model-based, offline and curriculum RL on thousands of parallelised simulations to shrink sim-to-real gaps.
4. Diffusion Policies & Generative Control. Leverage conditional denoising diffusion to produce temporally coherent whole-body trajectories for manipulation and locomotion
5. Graphics & Physics Simulation. Extend high-throughput GPU simulators such as Isaac Gym for photoreal perception and contact-rich dynamics.
6. Whole-Body Motion Control. Combine optimisation, MPC and topology-aware planners with learned priors to achieve robust running, climbing and bimanual dexterity.
7. Robot Morphology & Hardware-Aware Learning. Co-design actuation and policy to maximise energy efficiency and task transfer.
8. System-Level Integration. Build continuous deployment pipelines that stream policies from cloud training clusters to heterogeneous fleets in minutes.

Key Responsibilities

• Propose, implement and lead independent research projects in one or more focus areas.
• Publish in premier venues (Science Robotics, Nature Machine Intelligence, NeurIPS, ICRA) and release reproducible open-source code.
• Architect large-scale data pipelines that mix human video, synthetic worlds and on-robot logs to pretrain foundation models.
• Collaborate with hardware, simulation and DevOps engineers to validate algorithms on photoreal simulators and physical platforms.
• Mentor graduate interns and contribute to an inclusive, high-energy lab culture.

Minimum Qualifications

• PhD (awarded within the last five years) in Computer Science, Robotics, Machine Learning, Control, Graphics, Neuroscience or a related discipline.
• At least one first-author paper in a Tier-1 venue on embodied AI, multimodal learning, RL/IL, simulation, robot control or generative models.
• Fluency in Python/C++, deep-learning frameworks (PyTorch or JAX) and modern simulation stacks.
• Proven ability to work in interdisciplinary teams and communicate research clearly in English.

Preferred Qualifications

• Hands-on experience training billion-parameter transformers or diffusion models on multi-node GPU clusters.
• Contributions to public ML/robotics libraries.
• Prior work on world-model-based planning, differentiable simulators or multi-agent coordination.
• Hardware skills in mechatronics, perception or embedded real-time control.
• Track record of transitioning research prototypes into industrial pilots, mirroring recent factory deployments of humanoid robots trained via simulation-first RL