Staff Research Scientist, Reinforce Learning

Overview

As a Research Scientist at Wayve Labs, you contribute to the next generation of AI systems for autonomous driving. You will work at the crossroads of machine learning, simulation, robotics, and real-world deployment to push embodied AI frontiers. The role emphasizes multi-year breakthroughs within a high-conviction research team, balancing novel theory with scalable, real-world impact. You will collaborate across disciplines to tackle world modeling, multimodal learning, and scalable decision-making at scale.

Pay / Benefits

• relocation support with visa sponsorship
• flexible working hours
• private health insurance
• daily yoga
• onsite chef
• enhanced parental leave

Responsibilities

• Develop World Models and Planners for realistic simulation (diffusion-based, autoregressive, or hybrid)
• Advance reinforcement learning and reward modeling across real and synthetic data
• Develop Geometric Foundation Models for 3D spatial understanding
• Enable Cross-Embodiment Robotics using multimodal foundation models
• Conduct empirical research on scaling laws, generalisation, and sim-to-real transfer
• Define and evolve evaluation frameworks and benchmarks for long-horizon prediction and driving performance

Key requirements

• 3+ years of experience developing and deploying ML systems in real-world or production settings
• Deep expertise in one or more Embodied AI areas: foundation models, generative world modeling, reinforcement learning, or Spatial AI
• Track record of publications at top-tier conferences
• Strong programming skills in Python with PyTorch
• Data-centric mindset with experience on large-scale datasets and evaluation
• Strong problem-solving ability and collaboration in interdisciplinary teams

• collaboration in interdisciplinary teams
• strong problem-solving
• effective communication of complex ideas
• Foundation models (transformers, MoE)
• Generative world modeling (diffusion, autoregressive)
• Reinforcement learning (offline RL, RLHF, reward modeling)

…