SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Wayve is building the leading AI platform for autonomous driving, pioneering an end-to-end AI approach that enables vehicles to learn from real-world experience. The Measurement team within AI Evaluation builds and qualifies offline scene-understanding models used to measure driving performance after on-road runs and in simulation.
In this role, you will build, train, and fine-tune scene-understanding models for offline measurement, adapting on-vehicle architectures and Wayve foundation models for offline use. You'll improve accuracy and generalization across vehicle platforms, geographies, and driving conditions, benchmark models, diagnose failure modes, and use error analysis to guide iteration. You'll define quality bars and statistically credible evidence for validation pipelines and safety cases, collaborate with on-vehicle modelling, evaluation, data curation, and simulation teams, and mentor others while contributing to strong machine-learning and engineering practices.
You'll work on transformer-based, multimodal, and vision-language models for detection, segmentation, classification, and scene understanding. The focus is on offline models that leverage higher compute budgets, larger capacity, and bidirectional temporal context. You'll develop automated benchmarks that assess coverage, mine rare events, and measure driving behaviour, working with fleet-scale camera, lidar, and simulation datasets. Your models will generate evidence used in model-development decisions, validation pipelines, customer deliverables, and safety cases.
Wayve operates a hybrid model combining in-person collaboration with focused remote work. The company is not a mature, fully-structured organization—much of how it works is still being written, and you'll help shape it. This suits people who want real ownership over their work.
REQUIREMENTS:
- 4+ years in ML engineering with experience training and shipping deep-learning models in production
- Hands-on experience with modern computer-vision, transformer, multimodal, or VLM architectures
- Experience adapting or fine-tuning large pretrained or foundation models, with understanding of data, loss, and representation trade-offs
- Proficiency in Python and PyTorch or similar framework; comfort with large-scale training
- Ability to take ambiguous modelling problems from scoping to a working, measurable solution
- Valuable: experience in 3D scene understanding, offline modelling, auto-labelling, temporal or world models, autonomous vehicles, robotics, or distributed training
The company notes that not meeting every requirement is acceptable if you're passionate about autonomy and keen to learn.
