The future of AI won’t happen behind a screen.
We need embodied AI. Intelligence that can safely perceive, reason, and act in the physical world.
Wayve Labs is our frontier research unit, dedicated to defining this next chapter in machine learning.
Wayve Labs
Embodied AI for the Real World
Since Wayve was founded, research has been a critical part of how we stay ahead. Our work has spanned world models, vision-language-action models for driving, and beyond. We’ve shown that these are not just interesting ideas – they are the keys to unlocking complex physical tasks that can transform the way we live, move, and interact.
Our Vision
Wayve was founded nearly a decade ago on the premise that AI will have its greatest impact in the real-world.
Wayve Labs was born out of this vision. We believe that truly solving embodied AI requires creating the best possible home for scientists who want to solve one of AI’s hardest problems. A place with scalable data and compute resources, and a research team with the creativity and determination to build the foundations for the next stage of AI.
Our Focus
Our research spans the most challenging problems in the field today, focused on solutions that scale to the unstructured complexity of the real world.
Representation Learning
How do we learn representations that efficiently preserve the features of the multi-modal input signal that are salient for decision-making (agents, semantics, and beyond), while discarding the irrelevant clutter?
Spatio-Physical Intelligence
How do our models understand the spatial structure and physics of the dynamic 3D environments they operate in?
Decision-Making
How can we design an architecture and the policy learning algorithms that scale with data and compute? What role do long-term memory, language, and reasoning systems play? What should the interface between the learning algorithm and the robot be?
Learning Systems
How can we squeeze the greatest performance from a given corpus of data, or training/inference compute budget? In embodied AI this is critical for the proliferation of intelligence to any robot, anywhere.
Cross-Embodiment Learning
How can we use the data and systems we built for driving to accelerate deployment to diverse robotic platforms across mobility and manipulation?
World & Reward Modeling
How do we build a generative world model that is both diverse enough for exploration and reliable enough for closed-loop planning? How do we learn the reward model for safe and performant driving? Can we close the loop and learn in imagination?
Our Approach
We hunt for the big unlocks and the zero-to-one capabilities by empowering high-risk, high-reward research.
Wayve Labs is an ambitious, globally diverse, high-conviction research team. We have the strategic patience and commitment to prioritise multi-year breakthroughs over incremental gains.
Our History of Firsts
Our research team is constantly searching for things that slow us down, which has led to a rich, pioneering history of industry-firsts over the last decade.
Deep Learning
Deep Learning
Learning to Drive in a Day
We were the first to put deep reinforcement learning on an autonomous vehicle. In less than 20 minutes, we were able to teach a car to follow a lane from scratch.
World Models
World Models
Dreaming About Driving
We were the first to integrate a world model directly into a driving system. It took another five years for this work to have broader product impact.
Simulation
Simulation
Simulation Training, Real Driving
Our autonomous car drove on real UK roads by learning to drive solely in simulation, making us the first to test a model trained only in simulation in the real world.
GAIA
GAIA
Generative World Model
We were the first to build a generative AI foundation model built for driving, releasing GAIA to the world back in 2023.
LINGO
LINGO
Vision-Language-Action
We introduced LINGO, the first vision-language-action driving model to enhance how we interpret, explain and train our foundation driving models.
Our Breakthroughs
Research
LA-Pose: What 10 million Driving Videos Taught Us About Camera Pose Estimation
Research
GAIA-3: Scaling World Models to Power Safety and Evaluation
Research
Rig3R: Learning 3D Perception for Autonomous Vehicles
Research
Introducing PRISM-1: Photorealistic reconstruction in static and dynamic scenes
Press release
Driving with Language: Introducing Wayve’s Multimodal Driving Model LINGO-2
Research
GAIA-3: Scaling World Models to Power Safety and Evaluation
Research
Rig3R: Learning 3D Perception for Autonomous Vehicles
Research
Introducing PRISM-1: Photorealistic reconstruction in static and dynamic scenes
Press release
Driving with Language: Introducing Wayve’s Multimodal Driving Model LINGO-2
Publications
26 Mar 2025
GAIA-2: A Controllable Multi-View Generative World Model for Autonomous Driving
12 Mar 2025
SimLingo: Vision-Only Closed-Loop Autonomous Driving with Language-Action Alignment
13 Oct 2023
Driving with LLMs: Fusing Object-Level Vector Modality for Explainable Autonomous Driving
18 Oct 2021
FIERY: Future Instance Prediction in Bird’s-Eye View from Surround Monocular Cameras
07 May 2021
Video Class Agnostic Segmentation with Contrastive Learning for Autonomous Driving
Join Wayve Labs
Building the future of Embodied AI
Wayve was built on a contrarian, end-to-end AI approach to autonomous driving; Wayve Labs extends that mindset to the frontier of embodied AI, and we’re hiring. We’re looking for scientists and researchers who want to work on unsolved, high-impact problems, build systems that interact with and learn from the real world, and take ownership of bold ideas and see them through to real-world outcomes.
