← Company intelligence

Physical AI software · early

Perceptron AI

Corporate marketfoundation-modelsperceptionCurated record

Perceptron AI is a physical AI research lab co-founded by former Meta FAIR multimodal researchers, building foundation models for embodied perception and real-world reasoning in robot and edge environments. Its Mk1 model (launched May 2026) is purpose-built for video understanding and embodied reasoning: it processes native video at up to 2 fps across a 32K-token context window and benchmarks competitively with Anthropic, OpenAI, and Google frontier models at 80–90% lower cost. Target deployment environments include robotics perception pipelines, manufacturing quality control, geospatial analysis, and security monitoring; all domains requiring tight latency and cost constraints at the edge. BVP confirmed Perceptron as a portfolio company in its 2026 robotics and physical AI outlook alongside Waymo, ANYbotics, DroneDeploy, and Breaker.

HeadquartersSeattle, US
Founded2024
Corporate funding$16M
Team1-10
Deployments0
Open roles0
01

Company overview

Identity and operating footprint

Company typePhysical AI software
Market segmentCorporate
StageSeed
FoundersArmen Aghajanyan (Co-Founder & CEO; ex-Meta FAIR), Akshat Shrivastava (Co-Founder & CTO; ex-Meta FAIR)
Regions servedUS
Countries deployedNot listed
Service footprintNot listed
Last reviewed2026-07-05
CA

Funding history

Publicly announced rounds

Investors on recordBessemer Venture PartnersS32Foundation CapitalPlay TimeSmartGateVCThe Explorer Fund
TL

Company timeline

Milestones and announcements

2026-05-12Launched Mk1 physical AI foundation model for video understanding and embodied reasoning: native video at 2 fps, 32K-token context window, performance competitive with frontier labs (Anthropic, OpenAI, Google) at 80–90% lower cost; targets robotics, manufacturing, geospatial, and security deploymentsVentureBeat
2024-11-01Founded in Seattle, WA by Armen Aghajanyan and Akshat Shrivastava, former Meta FAIR research scientists specializing in multimodal foundation models; mission: build AI that perceives, reasons, and acts in the physical world reliably and at scaleBusinessWire
SC

Source ledger

2 unique public sources