Platform Architecture

Five layers. Zero cloud dependency. Complete sovereign control from input to inference.

How the modules work together

Every module is purpose-built. Together, they form complete operational pipelines.

Train and deploy a classified coding assistant

Pipeline: AriaOS Forge fine-tunes codebase, HammerIO compresses checkpoint, AriaOS audits inference, MemoryMap monitors memory

Secure perimeter monitoring: fully offline

Pipeline: AegisOS ingests camera feeds, AI Detection runs locally, HammerIO archives footage, AriaOS maintains audit log, SentinelForge correlates threats

Edge node deployment and recovery

Pipeline: AriaOS governs config, HammerIO compresses for transfer, node reboots, HammerIO restores in under 1 second, MemoryMap confirms memory state

Multi-jurisdiction emergency operations

Pipeline: AriaOS EOC ingests incident state, dual-tier AI drafts actions, HITL gate approves, Exchange Layer dispatches via EDXL-RM and CAP to jurisdiction platforms.

Validated hardware platforms

Tested under sustained load, chaos conditions, and air-gapped field deployment.

Platform	Memory	GPU Engine	Status
Jetson AGX Orin 64GB	64 GB	nvCOMP LZ4	Primary, fully validated
Jetson AGX Orin 32GB	32 GB	nvCOMP LZ4	Supported
Jetson Orin NX	16 GB	nvCOMP LZ4	Supported
Apple Silicon M-series	Shared	Metal / LZFSE	Supported
HP ProLiant G8	,	CPU	Validated (chaos testing)
Any CUDA GPU	,	nvCOMP LZ4	Supported
CPU-only	,	zstd / gzip / lz4	Full fallback

Deployment modes

One platform. Three operational postures. Deploy where you need it.

Field Deployment

Single Jetson node. Battery-powered optional. Air-gapped. All modules running. Validated at 99.97% uptime under load.

Base Installation

Rack-mounted or desktop. Multi-node capable. Web dashboards accessible on local network. Full telemetry and audit logging.

Development / Research

Workstation deployment for model training, benchmark validation, and capability development. Full pipeline from data to deployed model.