Key Notes
Embodied-AI research spans world and reward modelling, spatial intelligence, scalable decision systems and cross-embodiment learning for autonomous driving. Applicants need three-plus years developing and deploying real-world or production ML systems, a relevant postgraduate degree or equivalent experience, and Python with frameworks such as PyTorch. The full-time London role describes hybrid working and offers relocation support with visa sponsorship.
What You'll Work On
- Develop world models and planners using diffusion, autoregressive or hybrid approaches for realistic, consistent autonomous-driving simulation.
- Advance reinforcement learning and reward modelling across real and synthetic data, including scalable learning frameworks for embodied agents.
- Build geometric foundation models for dynamic 3D environments and multimodal systems that transfer learning across robotic platforms.
- Research scaling laws, generalisation and sim-to-real transfer through empirical studies.
- Define evaluation frameworks for long-horizon prediction, scene fidelity and driving performance.
Why This Role Matters
- Wayve’s driving simulations use world models to represent the consequences and costs of actions, allowing candidate behaviours to be examined in realistic, consistent environments.
- Wayve’s cross-embodiment robotics programme uses multimodal foundation models to accelerate learning on diverse platforms, extending the settings in which embodied-AI methods are developed.
What They Are Looking For
- Experience: Three-plus years developing and deploying ML systems in real-world or production settings is required.
- Qualification: A PhD, master’s degree or equivalent experience in ML, computer vision, robotics or a related field is required.
- Technical: Deep expertise in at least one embodied-AI area, such as foundation models, world modelling, reinforcement learning or spatial AI, is required.
- Technical: Strong Python programming, experience with frameworks such as PyTorch, and large-scale datasets and evaluation are required.
- Experience: A publication record at top-tier conferences, such as NeurIPS, ICML, ICLR, CVPR, ICCV or CoRL, is required.
- Preferred: Autonomous-driving, robotics, simulation, large-scale training, sim-to-real or open-source research-infrastructure experience is desirable.