§ 00  /  Filing No. HP-EDG-2026.04 Helsinki · Finland
Edge Autonomy Infrastructure
For Contested Operations

Decisions
at the Edge.
Silence on
the Wire.

EdgeOS converts raw sensor streams into mission-grade intelligence onboard — through fusion, semantic abstraction, and disciplined transmission. Built for environments where the cloud is not coming.

§ 01 / DOCTRINE Filed 2026·05 Class Public Pages 02 / 16

Today's architecture assumes connectivity. Tomorrow's battlefield denies it.

The Connected World Deprecated

A doctrine inherited
from peace.

  • Continuous connectivity is presumed
  • SATCOM is reliable and persistent
  • Cloud processing is always available
  • Upstream transmission flows uninterrupted
The Contested World Operating Reality

A reality designed
by adversaries.

  • SATCOM is bandwidth-constrained and costly
  • RF spectrum is jammed, denied, or compromised
  • EMCON discipline mandates radio silence
  • Cloud analytics fail in disconnected operations
The Insight
The bottleneck is not sensing.
It is decision-grade abstraction.
01
Object presence
02
Behavioral deviation
03
Threat indicators
04
Multi-sensor correlations
§ 02 / ARCHITECTURE Layer Edge Infrastructure Posture Non-displacing Pages 03 / 16

The missing layer between sensor and mission system.

HardpointAI sits between sensors and command. It does not replace mission systems, flight software, radar processing chains, or prime-owned IP — it makes them decisive when the link goes dark.

FIG. 02 · EDGEOS DATA PATH · CONTESTED-OPERATIONS PROFILE
REV / 2026.04
Sensor LayerIN
RADAR
EO / IR
RF / SIGINT
AIS / ADS-B
Custom payloads
Edge Autonomy LayerHARDPOINT
Perception
Ingestion + normalization
Fusion
Multi-sensor correlation
Ontology
Semantic abstraction
Comms
EMCON-aware transmission
MissionOUT
C2 Systems
Mission Computer
TAK / STANAG
Operator Display
Existing Prime IP
Non-displacing by design.   EdgeOS adapts to existing architecture — sensors retain their pipelines, mission systems retain their authority, primes retain their IP. The platform contributes only what is missing: the autonomous reasoning layer that turns raw streams into decisions when the network is denied.
§ 03 / SOFTWARE Stack EdgeOS Functional Modules Four Pages 05 / 16

EdgeOS — functional stack.

Four interlocking modules. Each addresses a specific failure mode of cloud-dependent C4ISR architectures.

i01 / 04
EdgeOS / Perception

From signal to substrate.

  • Sensor data ingestion across heterogeneous payloads
  • Adapter framework for radar, EO/IR, RF, AIS, custom
  • Normalization across timing, frame, and coordinate systems
ii02 / 04
EdgeOS / Fusion

From substrate to track.

  • Multi-sensor correlation in real time, onboard
  • Track reconciliation and confidence scoring
  • Anomaly detection across spatial and behavioral dimensions
iii03 / 04
EdgeOS / Ontology

From track to meaning.

  • Cross-sensor semantic mapping into mission objects
  • Mission-level abstraction layer over raw data
  • Unified representation for downstream C2 consumption
iv04 / 04
EdgeOS / Comms

From meaning to transmission.

  • Policy-driven, EMCON-aware transmission control
  • Bandwidth prioritization tuned to mission posture
  • Burst transmission with deferred bulk transfer
Operational Intelligence Console EdgeOS v2.4 · Operator view
EdgeOS Operational Intelligence Console — live track map view
EdgeOS Operational Intelligence Console — detections, timeline and link health dashboard
§ 04 / HARDWARE Models HP-FN1 · FN1 Micro Class Forward Edge Node Pages 07 / 16

Self-sustaining compute at the forward edge.

HP-FN1 edge node deployed in its ruggedized transit case in the field
HP-FN1 · Field deployment
HP-FN1 Micro mounted on an unmanned platform
HP-FN1 Micro · Platform-mounted
Model HP-FN1

A node that brings its own infrastructure.

Ruggedized, self-contained, deployable in under a minute. The HP-FN1 carries the full EdgeOS stack with no dependence on external power, network, or fixed installation.

Compute
Up to 2000 TOPS onboard AI inferenceEdge-optimized SoC · runs full EdgeOS stack natively
Storage
12 TB encrypted NVMeHigh-speed array · offline-first data retention
Power
Integrated LiFePO₄ batterySolar / vehicle / shore-power compatible · extended autonomous runtime
Network
10 GbE · MANET mesh · 5G · SATCOM (VSAT / LEO)Multi-link redundancy with EMCON-aware routing
Form Factor
MIL-SPEC Pelican-class transit case · IP67Tool-free deployment · standalone or rack-mounted · air / ground / maritime
Variants
HP-FN1 · HP-FN1 MicroMicro: same EdgeOS stack in a compact SWaP-optimized package — smaller platforms, dismounted operations, dense meshes
MODE / 01
Standalone

Battery-powered, fully autonomous, air-gap capable. The HP-FN1 operates without any external infrastructure for the full mission window.

MODE / 02
Networked

Multi-node MANET mesh distributes inference across forward positions, with resilient routing under jamming and node loss.

MODE / 03
Gateway

SATCOM and 5G uplink relay EdgeOS-curated mission objects to C2 — only what matters, only when it matters.

No external power. No fixed infrastructure. No compromise.

§ 05 / FOUNDRY Loop Field → Factory Cycle Continuous Pages 08 / 16

Fielded systems that improve with use.

Traditional defense systems are frozen at delivery. Foundry closes the loop: operational data collected at the edge feeds model retraining, and improved models return to every deployed node as cryptographically signed updates — over any available link, or by physical media when there is none.

i01 / 03
Foundry / Collect

From theatre to dataset.

  • Deployed nodes capture the operational data that matters
  • Edge cases, environments, and signatures no lab dataset contains
  • Curated onboard — only training-relevant data moves
ii02 / 03
Foundry / Retrain

From dataset to model.

  • Field data feeds model retraining and evaluation
  • Perception and fusion hardened against real operational conditions
  • Every deployment makes the next one sharper
iii03 / 03
Foundry / Redeploy

From model to fleet.

  • Improved models pushed to every node as signed updates
  • Delivered over available links — or by physical media
  • Capability ships to the fleet you already own
§ 06 / ONTOLOGY Output Mission Objects Compression ≈ 99.7 % Pages 10 / 16

From sensor outputs to mission objects.

Ontology is not a label. It is a transformation: from a stream of disconnected detections into a coherent picture of what is happening — transmitted as meaning, not bytes.

Raw Sensor Output
RADAR
track_id: 381
Range 14.2 km · vel 24 m/s
EO / IR
bbox_conf: 0.82
Frame 002441 · class unconfirmed
RF / SIGINT
emitter detected
Band L · 1.275 GHz
FUSE · ABSTRACT
Mission Object
SURFACE VESSEL
Correlated · medium confidence
Heading 274° · trajectory consistent · class merchant
UNIDENTIFIED AIRBORNE
Radar + EO correlated
Trajectory anomalous · flagged for operator
MARITIME ANOMALY
Behavioral deviation
AIS silent · heading inconsistent with declared route
Ontology enables meaning-based transmission — not raw data streaming.
§ 07 / EMCON Posture Silent-by-default Profile Mission-policy Pages 12 / 16

Designed for contested environments.

EMCON is not a feature — it is the operating discipline of the platform. Transmission is treated as a precious, observable act, governed by mission policy.

  • Full onboard correlation during radio silence
  • Event-driven transmission instead of continuous streaming
  • Burst-mode uplink for mission-critical objects only
  • Deferred bulk transfer when bandwidth permits
  • Policy-based prioritization tuned per mission posture
  • Cryptographic envelope on every emitted packet
Airborne ISR profile.
  • Radar and EO fused onboard in real time
  • Raw video stored locally — never transmitted
  • Only correlated mission objects sent via SATCOM
  • Full sensor archive offloaded post-mission, on-net
§ 08 / APPLICATIONS Domains Air · Sea · Land Status Pilot-ready Pages 13 / 16

Where the layer earns its position.

i.
Airborne ISR Platforms
  • Onboard radar + EO correlation in flight
  • SATCOM bandwidth load reduced by orders of magnitude
  • EMCON-compliant by default
  • Event-driven uplink to C2
ii.
Maritime Patrol & Naval Platforms
  • Radar + AIS + EO fusion for surface picture
  • Vessel anomaly detection against route baselines
  • Reduced operator overload in long-duration patrols
  • Autonomous picture compilation under EMCON
iii.
Forward-Deployed Tactical Nodes
  • Multi-sensor aggregation at the forward edge
  • Local AI inference without reach-back
  • Distributed edge autonomy across the mesh
  • Tactical network resilience under jamming
§ 09 / INTEGRATION Posture Adaptive Standards STANAG · TAK Pages 14 / 16

EdgeOS adapts to existing architecture — not the other way around.

01 / Hardware Deployment
Where the compute lives.
  • HP-FN1 rugged edge appliance
  • Embedded module in existing mission rack
  • Custom integration with approved compute hardware
02 / Software Deployment
How the stack is delivered.
  • Containerized deployment
  • Modular microservices architecture
  • Sensor adapter plugin framework
03 / Interfaces
How it speaks to your stack.
  • REST and gRPC endpoints
  • JSON-based, schema-versioned exchange
  • TAK-compatible event export, STANAG-aligned mapping
  • Custom adapters for prime-owned mission systems

Secure-by-design.

§ 10 · CYBERSECURITY POSTURE
Secure boot chain
Disk encryption
Signed software updates
Role-based access control
Audit logging
Hardened Linux baseline
Air-gapped operation mode
Offline-first system design
Mutual TLS between services
Zero-trust service identity
Encrypted data in transit
Signed model updates
§ 11 / POSITION Thesis Infrastructure Origin EU sovereign Pages 15 / 16

We are not building a feature.
We are building the missing
infrastructure layer
of the
tactical edge.

01
Edge autonomy is inevitable.
02
Disconnected operations are the new normal.
03
Bandwidth constraints are permanent, not transitory.
04
HardpointAI becomes the control plane of tactical edge intelligence.
§ 12 / DISPATCHES Type Programs · Partners Updated 2026·07 Pages 16 / 16

Signals worth transmitting.

Program · 2026·06 Cohort 002 · Completed
Palantir

Palantir Startup Fellowship
completed.

HardpointAI has completed Palantir’s Startup Fellowship — Cohort 002, closing the program at its final demo event in June 2026. The fellowship connected a selected group of defense and critical-infrastructure startups with Palantir’s engineering and deployment ecosystem as we scale EdgeOS toward production.

Partnership Kinetic Mesh
Rajant

Rajant partnership:
EdgeOS on Kinetic Mesh.

HardpointAI is an authorized Rajant Kinetic Mesh Partner — reselling and deploying Rajant’s battle-proven BreadCrumb® MANET networking, which EdgeOS nodes operate alongside for self-healing meshes that stay connected through jamming, movement, and node loss.

§ 13 · BRIEFING

Deploy semantic intelligence at the edge.

Request a technical workshop or architecture briefing to explore integration, interface mapping, and pilot deployment pathways.

contact@hardpointos.com

Palantir Startup Fellowship — Cohort 002  ·  Rajant Kinetic Mesh Partner