Principal Research Scientist – Multimodal Reasoning & Embodied AI

Job description

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 Labs Team Wayve Labs develops the next generation of learning capabilities for embodied AI, using autonomous driving as a demanding real-world testbed. Within FORGE, our Language & Reasoning team explores how LLMs, VLMs and VLAs can connect world knowledge, grounded reasoning, memory and planning to physical action. We work across model architectures, training and post-training, with the goal of helping driving agents generalise to new environments, follow goals and constraints, and handle rare situations. The team collaborates across London and Sunnyvale, with opportunities to publish fundamental research. 🧠 Your day-to-day You will set and lead an ambitious research agenda at the intersection of LLMs, VLMs, VLAs and embodied AI, asking how agents can use world knowledge, visual understanding, memory and reasoning to make better decisions in the physical world. You will remain hands-on: developing new architectures and learning methods, training and evaluating large-scale models, and testing whether better reasoning leads to better behaviour in challenging driving scenarios. As a Principal Research Scientist, you will guide the technical direction of the team, mentor researchers, build strong collaborations with Engineering, and turn original ideas into rigorous evidence, influential publications and capabilities that shape Wayve's next generation of AI. 🧩 What you’ll be working on: - Advance multimodal foundation models: Research and develop LLMs, VLMs and VLAs that integrate visual understanding, language, reasoning and action for embodied decision-making. - Develop reasoning and planning capabilities: Explore structured and latent reasoning, long-horizon planning, temporal understanding, memory, self-reflection and test-time computation for physically grounded agents. - Push the boundaries of model training: Develop scalable training and post-training methods, including supervised fine-tuning, reinforcement learning, reward modelling, distillation and learning from human or AI-generated feedback. - Build steerable and adaptable policies: Enable embodied agents to understand instructions, follow goals and preferences, reason about constraints and generalise to novel environments and situations. - Explore agentic and hierarchical architectures: Investigate how deliberative reasoning, memory, external tools and fast reactive policies can work together, balancing intelligence, reliability and real-time constraints. - Bridge reasoning and action: Develop approaches that turn semantic understanding and reasoning into meaningful improvements in physical-world behaviour, rather than treating language generation as an end in itself. - Drive research from idea to evidence: Design rigorous experiments, build prototypes, establish challenging evaluations and demonstrate improvements in real-world capabilities. - Shape the research direction: Work closely with researchers across foundation models, representation learning, world modelling and reinforcement learning. Contribute original ideas, mentor others and help define the next generation of embodied AI systems. 🙌 You should apply if: - A strong research track record in deep learning, particularly in one or more of LLMs, VLMs, VLAs, multimodal foundation models or embodied AI. - Deep technical understanding of modern foundation model architectures, transformer-based models, multimodal representation learning and generative modelling. - Hands-on experience developing, training, fine-tuning or post-training large-scale models, with a solid understanding of the challenges involved. - Strong knowledge of modern learning methods, such as instruction tuning, reinforcement learning, preference optimisation, reward modelling or reasoning-oriented post-training. - Excellent understanding of machine learning fundamentals, combined with the ability to challenge existing approaches and develop novel research ideas. - Strong practical skills in Python, PyTorch and large-scale machine learning experimentation. - Demonstrated ability to independently lead ambitious research projects, from identifying important problems to developing solutions and validating their impact. - A PhD in Machine Learning, Computer Science, Robotics or a related field, or equivalent research experience, with a strong publication record or evidence of impactful research contributions. 🌱 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. More about Wayve: 🚀 Wayve is building the leading AI platform for au