Skip to content

About us

We’re a foundation model company, building physical intelligence

We're a small, highly-motivated team of researchers, engineers, and product builders developing multimodal intelligence that processes video, audio, text, and sensor data in real-time. We support customers through the full deployment cycle, from initial integration to production.

Perceptron team at work
Perceptron team outdoors
Robotics development at Perceptron

Where we started

Armen Aghajanyan and Akshat Shrivastava previously led multimodal research initiatives at FAIR (Facebook's AI research team). In November 2024, they left to build a family of models designed specifically for the physical world: models that can perceive, reason, and act in any environment and accomplish any task.

Our model family

Mk1.5

Closed source

Our latest flagship: a vision-language model for image, audio and video understanding, with frontier embodied reasoning and video tracking.

Mk1

Closed source

A vision-language model (VLM) that is frontier on embodied reasoning and video understanding

Isaac 0.1

Open source

A vision-language model (VLM) that hit frontier on image grounding and localization

Isaac 0.2

Open source

A vision-language model (VLM) that hit frontier on image-reasoning via thinking-traces and tool calls

Isaac 0.5

Open source

An embodied foundation model that unifies video understanding, embodied reasoning, and robot control

Explore open roles

Physical environments generate massive streams of multimodal data that current AI can't process effectively. We're creating the intelligence layer that makes sense of this complexity: AI that adapts across hardware, scales across industries, and performs reliably in unpredictable conditions. Our systems are deployed with enterprise customers in manufacturing, media, logistics, and security, with a rapidly expanding pipeline. If you’re interested, ping us at join-us@perceptron.inc or apply through the open roles below.

Research

Engineering

Other