Machine Learning Engineer, Synthetic Data

Overview

You will advance GAIA-class world models and the synthetic-data pipeline to accelerate autonomous driving research at Wayve. You’ll bridge research and engineering, scaling generation and training to deliver training-ready artefacts. Your work spans post-training, failure diagnosis, and bringing synthetic data into real-model training, enabling faster iteration and deployment. You join a cross-functional team focused on practical, scalable AI that improves driving performance across new vehicle platforms. This role offers ownership of a core capability that directly impacts how quickly Wayve can train and evaluate driving models.

Pay / Benefits

• salaries benchmarked against market
• meaningful equity
• relocation support and visa sponsorship
• hybrid working with core hours
• learning and development budgets
• comprehensive benefits including health/dental

Responsibilities

• Post-train or ablate GAIA rig-transfer or pose-transfer checkpoints and diagnose failures in outputs (e.g., odometry bias, flickering)
• Manage generation at scale by launching tens of thousands of data segments and debugging GPU capacity and related parameters
• Land synthetic data into training pipelines, including binarisation and sampling hooks for RL and on-road differences
• Expand capabilities with new transfer sets and OEM rigs for broader coverage and safety scenarios
• Collaborate with world-model researchers, infra, and driving-model owners to ensure generation, evaluation, and training form a coherent system
• Alignment with research and publishing areas to ensure changes in checkpoints or data mixes translate to training impact

Key requirements

• 4+ years in applied ML / research engineering with proven training and shipping of neural nets
• Strong Python and PyTorch (or equivalent); comfortable with GPU training and reading model code
• Hands-on experience with video, generative or world models (diffusion/flow-matching/autoregressive video, novel-view synthesis, neural rendering)
• Working knowledge of cameras and 3D geometry (multi-camera rigs, intrinsics/extrinsics, warps/reprojection) and why they affect generation or training
• Evidence of using generated/simulated data to improve downstream models and measure impact (mix, ablations, failure analysis)
• Ability to operate large-scale generation or training (multi-GPU, workflow orchestration) and make it reliable
• Collaborative, experimental mindset; capable of owning a capability rather than a ticket
• Desire to learn and contribute even if not meeting every requirement

• collaborative
• experimental
• ownership
• Python
• PyTorch
• GPU training and debugging

…