Key Notes
Five years’ ML engineering or applied research and four years’ people management are required for this Wayve lead role. You’ll turn GAIA world models into training-grade synthetic driving data, owning generation, evaluation and downstream-training integration. Hands-on temporal generative models, camera geometry, Python and PyTorch are required, with evidence that synthetic data improved a trained model. The full-time London role is hybrid and combines technical design with direct-report development.
What You'll Work On
- Set synthetic-data technical direction, developing world-model conditioning for camera-rig transfer, pose transfer and controlled training-grade driving scenarios.
- Connect generation, evaluation and driving-model training, maintaining reproducible lineage from checkpoints through large-scale GPU inference to usable training artefacts.
- Improve generation throughput and usable yield through inference optimisation and self-service workflows, coordinating researchers, platform engineers and driving-model owners.
- Lead and grow synthetic-data engineers and scientists, managing quarterly priorities, experiments, feedback and team capability against safety and OEM programme requirements.
What They Are Looking For
- Experience: At least five years’ ML engineering or applied research with neural-network training and deployment, plus four years’ people management.
- Technical: Deep temporal generative modelling and hands-on video or world models, including diffusion, flow matching, autoregressive methods or VAEs.
- Domain: Camera and 3D geometry, including rigs, intrinsics, extrinsics, warps and reprojection.
- Experience: Evidence of generated data improving a downstream trained model, using ablations, mix ratios or failure analysis.
- Technical: Large-scale multi-GPU generation or training and strong Python and PyTorch engineering for production research tools.
- Stakeholder: Clear communication, mentoring and cross-functional project ownership.