Before the detail, here's the challenge you'd help us solve.

We build the embodied intelligence that moves real vehicles safely, and the ecosystem a billion machines will run on in the future. Very few people in AI can say this. Every role here, whatever the team, plugs into that.

Here's what this particular role covers.

About our Science Teams

We are looking for a Research Scientist to join the Multi-Embodiment Generalist Agent (MEGA) team within Wayve Science as a founding member. MEGA is building foundation models for general-purpose robots: models that learn from large-scale video, language, and robot-interaction data, then generalize across tasks and embodiments, including mobile manipulators, dual-arm platforms, and humanoids. Our aim is to build agents that can perceive, reason about, and act reliably in the physical world. You will help define and build the foundation-model learning stack for robotics: novel model architectures, pre-training objectives, post-training methods, and scalable data and training systems. The work combines frontier ML research with a direct route to real-world evaluation on a growing fleet of robots. Your work may span vision-language-action models, world and action models, video and multimodal models, imitation learning, reinforcement learning, and self-supervised learning. You will work with large-scale video and robotics datasets and distributed training infrastructure to develop increasingly capable, robust, and general robot policies. You will collaborate with research scientists, ML engineers, roboticists, and hardware teams to turn promising ideas into large-scale experiments, strong research contributions, and compelling robot demonstrations.

Your day-to-day

• Design and run experiments on model architectures, learning objectives and data strategies for robot foundation models.
• Build scalable pre-training and post-training methods using web video, egocentric video and robot-interaction data.
• Curate, filter, mix and evaluate large robotics datasets, including egocentric, UMI and teleoperated data.
• Develop and use distributed training pipelines for large multimodal models and datasets.
• Partner with ML engineers, roboticists and hardware teams to turn research progress into measurable real-world robot performance.
• Communicate findings through rigorous internal reviews, publications and robot demonstrations.

What you'll be working on

• Research and develop model architectures, learning objectives, and data strategies for robot foundation models.
• Develop scalable self-supervised and generative pre-training methods using web video, egocentric video, and robot-interaction data.
• Develop post-training approaches, including supervised fine-tuning, imitation learning, reinforcement learning, and related methods, to improve real-world robot capabilities.
• Curate, filter, and evaluate large-scale robotics datasets, including egocentric, UMI, and teleoperated data.
• Build and use distributed training pipelines for large models and large multimodal datasets.
• Work closely with robotics and hardware teams to connect model progress to measurable real-world performance.
• Communicate research clearly internally and, where appropriate, through publications and Wayve's scientific presence.

You should apply if

• Experience in machine learning, with a focus on multimodal foundation models and data for foundation models.
• Experience with scalable training, such as multi-node training, large datasets and/or large model training.
• Strong research track record, including publications in top-tier venues such as ICRA, CoRL, CVPR, NeurIPS, ICML or ICLR.
• Strong engineering skills and hands-on experience with modern machine learning frameworks.
• Ability to design and run rigorous experiments while collaborating closely with engineering and robotics teams.
• Experience translating research ideas into working systems, experiments or deployed capabilities.
• Strong communication skills and the ability to share research clearly across teams.

Not ticking every box? That's totally okay! If you're passionate about autonomy and keen to learn, we encourage you to apply even if you don't meet every requirement.

How we work - Locations & Flexible Working:

Our main hubs are in London, Sunnyvale, Yokohama, Herzliya, Vancouver and Leonberg. We operate a hybrid working model that combines in-person collaboration in our dedicated office spaces with focused time working remotely. This gives our teams the connection and energy of working together, alongside the flexibility to do their best work in a way that fits their lives.

The Interview Process:

Our process is clear and respectful of your time:

• Initial call / recruiter screen (30 mins)
• Competency Interviews (Pytorch debugging and hiring manager interview; 1.5 hours total)
• Deep-dive technical interviews (Systems & domain-specific interviews; 3 hours total)
• Final interview: Mission & values alignment (45 mins).

We'll always explain the format and work around your availability.

What's in it for you (Location dependant):

Salaries benchmarked against the market annually
Meaningful equity, sharing in the ownership and long term success of Wayve
Relocation support and visa sponsorship where applicable
• Hybrid working, core hours and the chance to work hands on in vehicle workshops and labs
Learning and development budgets with support for training, conferences and growth
Comprehensive benefits including health insurance, dental, enhanced maternity and paternity leave, retirement or pension where applicable, access to therapists, wellbeing partnerships, team socials and more

A quick, honest note before you apply.

Wayve is not a mature, fully-structured place with the playbook already written. Much of how we work is still being written, and if you join, you'll help write it. That suits people who want real ownership more than people who need a settled structure from day one.

If that sounds like the kind of problem you want to spend your time on, we'd really like to hear from you.