Key Notes
Closed-loop simulation work centres on next-generation generative world models that can train and evaluate autonomous-driving systems interactively rather than merely replay recorded logs. Relevant expertise includes 4+ years in ML research or engineering, deep diffusion and latent-video model knowledge, temporal or spatial-temporal data experience, and strong Python and PyTorch. The role is full time in London under Wayve’s hybrid working policy.
What You'll Work On
- Develop efficient generative world models using diffusion, transformer or hybrid architectures for real-time autonomous-driving simulation.
- Build interactive simulation models that allow agents or humans to step through environments for reinforcement learning, planning and safety evaluation.
- Optimise inference latency and compute efficiency through model compression, context pruning and related performance techniques.
- Define metrics for long-horizon coherence, physics fidelity and planner integration, using ablations and scaling studies to test trade-offs.
- Integrate simulation models into closed-loop training and evaluation workflows and compare simulated outcomes with on-road driving-model results.
Why This Role Matters
- Wayve’s closed-loop training and evaluation stack depends on world models that can generate interactive driving environments fast enough for large-scale use, making model efficiency central to how much simulated experience can be produced.
- Autonomous-driving model evaluation relies on simulation that preserves long-horizon behaviour and physical fidelity, so model quality determines how confidently simulated results can be compared with on-road performance.
What They Are Looking For
- Experience: Relevant experience includes 4+ years in ML research or engineering focused on generative video or world models.
- Technical: Deep knowledge of diffusion and latent-video models, including improving sampling efficiency or model throughput, is sought.
- Technical: Experience with high-dimensional temporal or spatial-temporal data such as video or multi-sensor fusion is sought.
- Technical: Strong Python and PyTorch engineering fundamentals and experience building research-grade production tools are sought.
- Experience: A strong publication record or contributions to open-source ML tooling are sought.
- Preferred: Autonomous vehicles, robotics, simulation, embodied AI and synthetic-to-real transfer experience are desirable.