We are a London-based AI and Robotics research lab.
At Pixelwise AI, despite being scientists, we are also strong believers in markets. We are excited by complex, challenging, and non-obvious technical problems that create value for humanity when deployed at scale. We start with real-world, large-scale problems and reason backward to research hypotheses—not the other way around.
Robotics has a data problem: unlike language, there is no internet-scale dataset we can directly train robots on except for internet videos.
We believe that with advances in pixel-wise 3D computer vision, robots can be trained from internet human data, and we are on a mission to unlock this.
Previously, we worked on projects such as CoTracker, VGGT, Text-to-4D, Meta 3D Gen, and VGGSfM at Meta AI and the University of Oxford.
If you're excited about robotics and this vision—reach out to us at hi@pixelwise.ai