TensorWorld
We create frontier models that faithfully understand, simulate, and interact with the physical world.
TAP TO WAKE THE ARM
Physical AI promises intelligence that can understand the physical world, simulate how it behaves, and act within it. It is easy to be captivated by what AI can do. Yet fulfilling that promise requires us to pay as much attention to the world AI encounters as we do to AI itself.
The physical world follows its own laws, no matter how convincing a generated scene, an explanation, or a plan may be. A model can devise a plan that is sound in a world of its own imagining yet fails in the real one. Ultimately, the physical world determines the outcome.
Fidelity to physical reality cannot be taken for granted. The laws that govern the physical world must be integral to how a model interprets observations, predicts outcomes, and acts. At TensorWorld, we believe physical intelligence rests on making faithfulness to physical reality an intrinsic property of the model itself.
01
Video models generate scenes. Conventional simulators let us build virtual worlds in code. Yet both can only approximate physical reality. We seek models that can infer from observations the physical structure and dynamics that produced them and generalize beyond the assumptions built into simulators.
The world behind its representationsWe are building foundation models that internalize the structure and dynamics of the physical world. Our goal is to make faithfulness to physical reality intrinsic to these models—shaping how they understand the world, predict its behavior, and act within it.
With these capabilities, we aim to empower robots to interact reliably with their surroundings, help engineers create better designs, and enable scientists to make new discoveries.
We want these models to carry physical knowledge across tasks and environments: reasoning about unfamiliar situations, anticipating the consequences of actions, and learning from new evidence.
RoboticsComputingEngineeringIndustryDiscovery
RoboticsComputingEngineeringIndustryDiscovery
We have helped shape how machines understand the world—its physical laws, temporal dynamics, and three-dimensional structure.
We advanced physics foundation models with Transolver and GeoPT, bringing the principles of physical systems into pre-training. We developed Autoformer and TimesNet to uncover structure across time, and scaled up 3D representation learning with Uni3D. We pushed the frontiers of learning from limited experience with EfficientZero, enabling agents to plan through learned models of the world.
In previous roles, we led the development of products serving more than 100 million users and oversaw computing infrastructure spanning 100,000 accelerators. At TensorWorld, we draw on that experience to build frontier models at scale and turn research into products.
Today, we unite these strengths around a shared ambition: intelligence that understands how the world actually works, and acts reliably on that understanding.
