US AI startup building multimodal foundation models for physical AI — perception, reasoning and robot control. Creator of the Isaac model family and Perceptron Mk1.
Perceptron (Perceptron Inc.) is a US artificial-intelligence startup building multimodal foundation models for so-called physical AI — systems that process video, audio, text and sensor data in real time to perceive, reason and act in the real world. The company was founded in November 2024 by Armen Aghajanyan and Akshat Shrivastava, who previously led multimodal research at Meta FAIR (Facebook AI Research). It is headquartered in Bellevue, Washington. Perceptron develops the open-weight Isaac model family (0.1, 0.2, 0.5) and the closed-source Perceptron Mk1, along with a platform and APIs for deploying perception. Isaac 0.5 is an open-weight, 36-billion-parameter sparse embodied foundation model that unifies video understanding, embodied reasoning and robot control. Its systems are deployed with customers in manufacturing, logistics, media and security. As of August 2026 Perceptron had raised about $21M from Bessemer Venture Partners, Foundation Capital, S32 and SmartGateVC.
Founders
Previously led multimodal research initiatives at Meta FAIR (Facebook AI Research). Co-founded Perceptron in November 2024.
Previously led multimodal research initiatives at Meta FAIR (Facebook AI Research). Co-founded Perceptron in November 2024.
Classification
External links
