The edge OS for physical AI.
PhyOS runs on edge gateways, robots, drones, kiosks, and screens. Intel, ARM, NVIDIA. It turns any supported device into a fleet-managed, AI-capable edge node. Container runtime with GPU acceleration, OTA updates with atomic rollback, and a self-healing watchdog that keeps devices running when the network doesn't.
Container-based deployment
Deploy apps as containers on edge hardware. Same tooling, same workflow, same images you use in the cloud.
GPU-accelerated inference
Run models with GPU acceleration across NVIDIA CUDA, AMD ROCm, and the Intel equivalent. The runtime handles GPU scheduling and memory management, and it is where models trained on your fleet telemetry deploy back to the edge.
OTA updates with rollback
Push updates across your fleet. If an update breaks, the device rolls back without intervention.
Offline-first operation
The edge keeps running when the cloud is unreachable. State syncs when connectivity returns.
Any form factor
Edge boxes, robots, drones, kiosks, screens. Same apps, same fleet management, same descriptors. Write once, deploy to any supported device.
The same app runs on an edge box, a robot, a drone, a kiosk, or a screen. It keeps deciding locally even when the network drops, and you deploy or roll back across the whole fleet from the console or API.
Deploy AI at the edge.
Today.
Flash PhyOS, provision the device, deploy your first app. The free tier includes 3 devices.