Senior ML Engineer, AI Performance & Edge Deployment

Wayve; Greater London; Hybrid; GBP 120,000 - 180,000; Full time

Benefits offered by this job

Relocation support

Hybrid working

Learning budget

Equity sharing

Health insurance

Dental benefits

Parental leave

Job summary

Wayve is building the leading AI platform for autonomous driving. We are seeking an ML Optimisation Engineer to own end-to-end model releases and drive performance improvements for on-vehicle deployments.

You will train PyTorch models, apply quantisation, distillation and low-rank methods, and collaborate with ML and performance teams to ship reliable models while balancing latency, memory and power constraints.

Qualifications

• Proven experience improving performance in production systems with tight constraints.
• Hands-on experience training and iterating on deep learning models in PyTorch, beyond high-level tooling.
• Familiar with model optimisation concepts such as quantisation and/or distillation.

Responsibilities

• Own end-to-end delivery of model releases from requirements through training, evaluation and deployment readiness.
• Train and iterate on PyTorch models with a hypothesis-driven approach.
• Debug model performance by identifying regressions and proposing fixes.
• Collaborate with adjacent ML and performance teams to hand off models and align on optimisation priorities.
• Communicate delivery timelines, trade-offs and readiness criteria to stakeholders.

Skills

PyTorch

Model optimization

Performance engineering

Kernel/runtime understanding

Edge deployment

Quantisation

Distillation

Embedded/edge deployment

Tools

TensorRT

CUDA

Qualcomm QNN

Triton

OpenCL

Job description

Wayve is building the leading AI platform for autonomous driving. We are seeking an ML Optimisation Engineer to own end-to-end model releases and drive performance improvements for on-vehicle deployments.

You will train PyTorch models, apply quantisation, distillation and low-rank methods, and collaborate with ML and performance teams to ship reliable models while balancing latency, memory and power constraints.