Perception
How can compact models turn sensor data into useful representations of their surroundings?
We research compact AI models for systems that sense and act in the physical world.
MOVE TO EXPLORE · TAP TO PULSEMachines operating on their own can't assume a reliable connection, ample compute, or unlimited power. We're exploring what compact models could make possible when intelligence has to work onboard, within the limits of the device.
Our primary interest is how people and AI can work together in constrained environments. These directions are a starting point, with room to follow new questions and use cases.
How can compact models turn sensor data into useful representations of their surroundings?
What capabilities can we preserve under strict memory, power, and compute limits?
How could perception and reasoning guide action in changing physical environments?
Smaller models.
Closer to the world.
Built for physical constraints.
We're exploring where compact, onboard intelligence could matter most. Have a use case, a research question, or a perspective we should hear? We'd like to talk.