Approach &Technology

Systems that decide
without the network

Technology

HardpointAI builds EdgeOS, the operating layer for systems that must decide without the network. It runs on the node, turns raw sensor streams into tracks and events, and sends only what matters, when it is allowed to. Next comes our own foundation model for the edge, and autonomy that can be proven. We develop and train it in-house, and build the nodes that carry it, in Vantaa, Finland.

Our approach

On the Node
01
TodayDetection, tracking and fusion run on the node itself, with the network switched off. A link, when there is one, carries sync, fleet learning and signed updates. It is never a dependency.
Many Sensors, One Picture
02
TodayCameras of every kind (IP, GMSL, infrared and thermal), drone video with its position data, radar and lidar, and other sensors through an adapter each. Built to see one person or vehicle in several sensors as one track.
Quiet by Policy
03
TodayThe unit decides what may leave the node, when, and over which link. Through an outage, results wait on the node and go when the link returns.
Our Own Foundation Model
04
BuildingDetection is plumbing. We are building a foundation model for the edge: one model that sees across cameras, drone video and other sensors, keeps one identity per real object, learns a new object type in hours, judges what is worth sending and knows when it is unsure.
Agents on the Node
05
NextAgents that reason over the fused picture and work through a fixed set of tools: task a sensor, cue another node, draft a report for the operator. Every call passes a policy gate.
Assured Autonomy
06
DirectionAutonomy that can be proven. The commander sets how far a system may act, mission by mission. Limits live in deterministic code outside the model, every decision is recorded and can be replayed, and every release ships with its test evidence from simulation and red-teaming. Today, EdgeOS advises and people decide.
Developed and Trained In-House
07
TodayEdgeOS and its models are developed and trained in-house, in Finland. They run on our own nodes: HP-FN1, with a 10.1-inch rugged touchscreen and up to 2070 TFLOPS on NVIDIA Jetson Thor, and the compact HP-FN1 Micro in its own enclosure. On a partner's platform, EdgeOS runs on the compute it already carries.

Sensors multiply, and the link is the first thing to go. We are building the layer that keeps a system deciding when it is gone: our own models on the node, quiet by policy, and autonomy that can be proven, with people in charge of what it may do.