01 / INF
Sovereign inference
The model, the manuals and the rules of engagement run inside the perimeter on the edge node — offline retrieval and on-node inference (llama.cpp / GGUF, vector search) with zero external calls.
Platform / Sovereign edge
The sovereign edge platform.
Autonomy, track fusion, threat intelligence and a common operating picture, unified into one foundation for national security — and engineered to run entirely inside your perimeter.
01 / Overview
One foundation, two operational systems.
Hyperborea builds two operational systems — edge autonomy and threat intelligence. Both rest on the same platform, so the model, the mission rules, the track picture and the protocol adapters are the same components, assembled differently. An integration completed once holds for both.
Every layer answers the same constraint: no cloud inside the trust boundary, and no assumption that a link will be available at the moment a decision has to be made.
- Runs on-node, inside the customer's own perimeter.
- Operates air-gapped, comms-denied and GPS-degraded.
- Ships as signed, reproducible artifacts with a full audit trace.
- Applications Edge autonomy and threat intelligence
- Intelligence layer On-node inference and signed decision logic
- Data fusion Track fusion and multi-source ingestion
- Edge infrastructure On-node compute and secure interoperability
02 / Architecture
Five layers, engineered for contested environments.
The stack reads bottom-up in deployment and top-down in use. Each layer is independently testable and carries its own assurance evidence, and nothing in it depends on a service we do not ship.
02 / DEC
Signed decision logic
Deterministic mission rules and playbooks as Ed25519-signed artifacts — no code execution, human-in-the-loop gates, and a comparison-by-comparison audit trace on every firing.
03 / COP
Common operating picture
Track fusion, threat evaluation and multi-INT resolved into one live operating picture — with engagement interlocks and full custody of every track.
04 / INT
Threat-intelligence ingestion
Multi-source OSINT and sensor feeds ingested, filtered, geolocated and corroborated into geofenced early warning — informal channels included.
05 / NET
Secure interop & mesh
Protocol adapters (Link-16, MIL-STD-1553, CoT/ATAK, STANAG) behind a cross-domain guard, with peer-to-peer sync between nodes — and no central server to lose.
03 / Core modules
Sense, know, act.
Three interoperable modules assemble into either system. They share the inference layer, the decision layer and the transport, which is what keeps the integration surface small.
Module 01Sense
Track fusion
Fuses radar, ADS-B, AIS, RF and reporting into one deduplicated track picture with threat evaluation.
- Radar · ADS-B · AIS · RF
- Deduplicated tracks
- Threat evaluation
Module 02Know
Threat intelligence
Ingests open, dark and technical sources, then corroborates and geolocates them into ranked warnings.
- Multi-source ingestion
- Cross-source corroboration
- Geofenced alerting
Module 03Act
Autonomy core
Runs signed, deterministic decision logic for edge and autonomous systems.
- GPS-denied navigation
- Auditable decisions
- Mission replanning
04 / Where it runs
Inside the perimeter, on hardware you control.
The platform runs on rugged edge compute at the site, the node or the vehicle — never in a vendor cloud. Nodes mesh peer-to-peer, and each one continues to operate independently when links drop, reconciling state on reconnect.
Review the platform with our engineers.
Technical evaluation, integration questions and field trials go directly to our engineering team.
Contact engineering