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 using end-to-end AI that enables vehicles to learn from real-world experience. The company is pioneering a mapless, hardware-agnostic approach that integrates with global OEM partners.
You will join the ML Optimisation Team (AI Performance), a high-ownership team responsible for delivering production-ready model releases as Wayve's OEM engagements and release cadence accelerate. The team is applied and delivery-focused, taking models from "works in training" to "meets product constraints" while working closely with downstream inference and performance specialists.
Day-to-day responsibilities include:
- Owning end-to-end delivery of model releases, from initial requirements through training, evaluation, iteration and deployment readiness
- Training and iterating on PyTorch models with a hypothesis-driven approach, running ablations against clear evaluation criteria
- Debugging model performance: identifying regressions, root-causing issues and proposing fixes
- Collaborating with adjacent ML and performance engineering teams to hand off models, define bottlenecks and align on optimisation priorities
- Communicating with stakeholders on delivery timelines, trade-offs and readiness criteria
Key focus areas:
- Getting models to meet tight runtime constraints on-vehicle as model capability grows
- Applying practical optimisation techniques such as quantisation, distillation and low-rank methods
- Shaping the handoff between training, evaluation and deployment for quick and reliable shipping
- Working at multiple levels of abstraction, from high-level model behaviour down to runtime and latency implications
Wayve operates a hybrid working model with main hubs in London, Sunnyvale, Yokohama, Herzliya, Vancouver and Leonberg, combining in-person collaboration with focused remote work.
REQUIREMENTS:
- Proven experience improving performance in production systems with tight constraints (latency, memory, bandwidth, power/thermal or cost)
- Strong hands-on experience training and iterating on deep learning models in PyTorch, beyond high-level tooling
- Proficiency with at least one relevant stack or toolchain (e.g. TensorRT, CUDA, Qualcomm QNN, Triton, OpenCL) and ability to pick up adjacent frameworks quickly
- Comfortable moving between high-level model behaviour and low-level kernel/runtime execution
- Familiarity with model optimisation concepts such as quantisation and/or distillation (hands-on experience is a strong signal, but solid fundamentals are sufficient)
- Strong engineering fundamentals and collaboration skills
- Bonus: experience with models under tight latency/efficiency constraints (edge, embedded, real-time), exposure to ML systems from training through to deployment handoff, and embedded/edge deployment including benchmarking on real devices
