Staff Machine Learning Engineer, Av Core

Wayve Wayve · Robotics · Sunnyvale, CA · AV Engineering

Staff Machine Learning Engineer on the Core Model Safety team at Wayve, focusing on developing and deploying end-to-end AV 2.0 models for assisted and automated driving. Responsibilities include driving the roadmap, training models on fleet data, building evaluations, and collaborating with research and engineering teams. Requires experience in ML engineering, Python/C++/CUDA, PyTorch, transformer/multimodal architectures (VLM/VLA), and multi-task training. Desirable experience includes AV/robotics, 3D scene understanding, reward modeling, interpretability, and safety-critical systems.

What you'd actually do

  1. Drive Core Model Safety roadmap themes owning the full lifecycle from research to offline/online experiments to technology transfer.
  2. Train and deploy end-to-end AV 2.0 models on our global fleet, using large-scale, diverse data to validate capabilities and improve generalisation across vehicles, markets, and driving conditions.
  3. Build high-value open-loop and closed-loop evaluations for core capabilities and representation learning.
  4. Align priorities and learn from the organisation - with AV Core, Evaluation, and Product Engineering on roadmaps and failure modes; from fleet, simulation, and product feedback; and through mentoring others on the team.
  5. Maintain awareness of the wider business context - division and company priorities, near-term product programmes, and how Core Model Safety work enables them.

Skills

Required

  • 5+ years in ML engineering
  • Python
  • C++
  • CUDA
  • PyTorch
  • software engineering practices
  • transformer-based architectures
  • multimodal architectures
  • vision-language models (VLM)
  • vision-language-action models (VLA)
  • training shared representations with multiple tasks or objectives
  • Staff-level technical leadership

Nice to have

  • autonomous vehicles
  • robotics
  • hands-on deployment
  • closed-loop validation on physical systems
  • 3D scene understanding
  • representation learning
  • geometric perception
  • semantic perception
  • large-scale semantic enrichments
  • reward modelling
  • behaviour modelling
  • model introspection
  • interpretability
  • redundant or fallback architectures
  • safety-critical systems
  • foundations/pretraining
  • applied engineering teams
  • large-scale training infrastructure
  • agentic workflows

What the JD emphasized

  • pathfinding in ambiguous problems
  • scoping and evals
  • establishing a direction
  • knowledge transfer
  • transformer-based and multimodal architectures
  • vision-language models (VLM)
  • vision-language-action models (VLA)
  • training shared representations with multiple tasks or objectives
  • real trade-offs across data and losses
  • Staff-level technical leadership
  • research-literate and pragmatic
  • setting direction
  • raising the bar
  • leading cross-functional work without formal line management
  • prior experience in autonomous vehicles or robotics
  • hands-on deployment and closed-loop validation on physical systems
  • 3D scene understanding
  • representation learning for geometric and semantic perception
  • large-scale semantic enrichments
  • reward modelling
  • behaviour modelling
  • model introspection
  • interpretability
  • redundant or fallback architectures
  • safety-critical systems
  • foundations/pretraining
  • applied engineering teams
  • large-scale training infrastructure
  • agentic workflows

Other signals

  • end-to-end driving model
  • trained capabilities
  • large-scale training and fleet data
  • foundational capabilities for assisted and automated driving
  • collision avoidance
  • scene understanding
  • model understanding
  • robustness under failure
  • transformer-based and multimodal architectures
  • vision-language models (VLM)
  • vision-language-action models (VLA)
  • training shared representations with multiple tasks or objectives
  • autonomous vehicles
  • robotics
  • 3D scene understanding
  • representation learning
  • reward modelling
  • behaviour modelling
  • model introspection
  • interpretability
  • redundant or fallback architectures
  • safety-critical systems
  • foundations/pretraining
  • applied engineering teams
  • large-scale training infrastructure
  • agentic workflows