GHOST
Tactical AI. On your device.
Ghost is an Android tactical application with the Orion language model embedded for fully local inference. It brings AI reasoning to the point of action without a cloud service or remote inference server.
Local reasoning. Connected reporting.
On-device inference
Orion runs directly on Android Ghost nodes, keeping local AI reasoning available when remote compute is unavailable.
Edge-to-headquarters exchange
Ghost nodes exchange data with Kinetic Defense through an existing backend gateway. Information can move to company headquarters and above when connectivity is available.
Human-supervised decisions
Ghost supports operator and staff review. Decision authority remains with people; no autonomous engagement or targeting authority is claimed.
From the operator to headquarters
In a dispersed company, Ghost nodes run Orion locally even when the route to headquarters is interrupted. When connectivity is available, tactical reports and relevant context pass through the backend gateway to Kinetic Defense for human-led planning and coordination.
Ghost + Orion
Backend gateway
Kinetic Defense
Local inference runs on the device. Headquarters data exchange requires connectivity.
Built today. Measured next.
Implemented baseline
On-device Orion inference and Ghost-to-Kinetic data exchange are implemented. The current prototype is at TRL 4, with testing conducted in laboratory and synthetic environments.
Next validation steps
Planned evaluations will measure device performance, interrupted-link recovery, message synchronization, security controls, and operator usability. Independent Army validation has not yet occurred.