A pocket-sized VAOS runtime that boots from USB-C and turns any host machine into a cognition node. Bring your own camera. Drop into any site. Validate, reason, act.
Camera / RTSP / Sensor
↓
VAOS Portable Runtime
↓
Observe → Structure → Think
↓
Validate → Act
↓
Dashboard / SDK / API / MQTT
Portable hardware. The same runtime that ships into production OEM modules — packaged for development, evaluation, and field deployment.
Build and test VAOS apps locally. Boot from USB on any laptop or workstation.
Add a cognition runtime to existing robots — perception, reasoning, validation, action.
Evaluate VAOS against your camera modules without committing to custom hardware.
Deploy local reasoning anywhere — field sites, vehicles, temporary installations.
Real-time cognitive overlays on existing RTSP/WebRTC feeds. Pair, plug, stream.
Same kernel, different targets. Pick by capability — not vendor lock-in.
| Edition | Target | Best for |
|---|---|---|
| VAOS Lite | Entry runtime | Lightweight, low-resource sites |
| VAOS Developer | SDK / testing | Build, debug, iterate locally |
| VAOS Stream | RTSP / live overlays | Real-time cognitive feeds |
| VAOS Edge | Robotics / OEM | Production edge deployments |
| VAOS Distributed | Federated / cloud-assisted | Multi-node cognition networks |
Not locked to CUDA. Not locked to cloud. Not locked to one vendor.
Rockchip — first-class NPU integration, hardware-accelerated ISP.
NVIDIA Orin Nano, Orin NX, AGX. CUDA-accelerated SceneLM variants.
Pi 5 + Hailo, Pi 4 baseline. Lite edition runs comfortably.
Intel / AMD with or without discrete GPU. Developer edition default.
UVC-class webcams, USB3 industrial. Standard pipeline.
Pull from any IP camera or stream. RTMP/WebRTC adapters available.
Hailo, Kinara, Cambricon, Axelera. Plugin-based silicon support.
Every Portable Runtime ships with a live status surface — pair from desktop, mobile, or browser.
Runtime Status ✔ Cognition Active ✔ Validation Enabled ✔ Memory Warm ✔ Sensor reliability 0.93 ✔ SceneLM Connected ✔ Throughput 28.4 FPS
Bridge VAOS into real-time control loops for high-determinism robotics.
Synchronised perception across stereo + wide + thermal + depth in one runtime.
SceneLM-predictive variant — proactive intervention from forward modelling.
Runtime auto-tunes ISP, exposure, and inference weights per stream condition.
Multiple portable runtimes federated into a single site cognition network.
Early access opens for developer and OEM partners. Hardware sampling underway.