Key Notes
Computer vision and 3D perception span Wayve’s ADAS model training, evaluation and dataset-generation work. Relevant expertise includes shipped CV deep-learning systems and applied 3D perception experience, while welcoming applicants who do not meet every item. The London remit may lean towards in-car models or offline data pipelines depending on your strengths, under Wayve’s hybrid working policy.
What You'll Work On
- Train, debug and improve computer-vision and 3D-perception models using evaluation signals to select the next ADAS performance gaps.
- Build scalable auto-labelling and pseudo-labelling pipelines that generate larger training datasets.
- Develop detection, classification and instance-segmentation capabilities for lanes, road objects, traffic signs and traffic lights across ADAS perception models.
- Contribute to tracking and 3D-reconstruction pipelines, or optimise online perception models for in-car latency and compute constraints.
Why This Role Matters
- Wayve’s ADAS perception stack identifies lanes, objects, signs and signals, so evaluation-led model iteration determines which observed driving-performance gaps are addressed.
- Wayve’s training datasets gain labelled coverage through offline tracking and 3D reconstruction, supplying model development with labels propagated through time.
What They Are Looking For
- Experience: Relevant experience includes work building and shipping computer-vision deep-learning systems beyond research-only work.
- Technical: Relevant technical skills include 3D-perception concepts or pipelines such as LiDAR, multi-view geometry, tracking or 3D reconstruction.
- Experience: Relevant experience includes end-to-end ownership of evaluation and dataset generation within applied machine-learning work.