System Operational · Dual-Use ISR

Onboard agentic triage
for bandwidth-constrained satellites & drones

Production-ready edge AI platform that filters and prioritizes satellite imagery before downlink — saving 85%+ bandwidth and surfacing items of interest in real time. CNN + ReAct agentic reasoning + HMAC audit trail. Built for NVIDIA Jetson Orin, under 20 watts.

Or clone & run locally: make install && make run

<20W
Power Envelope
>85%
Bandwidth Saved
<50ms
Per-Tile Latency
119
Tests Passing
14
Production Modules
Dashboard Preview

What you'll see in the live demo

Real-time mission-control interface — pulls live Sentinel-1 SAR, Sentinel-2, and NOAA GOES imagery. Runs ship detection, fire monitoring, and agentic triage in your browser.

interactiveintel-edge-ai-satellite-triage.streamlit.app
System Operational · Dual-Use ISR
Edge AI Satellite Triage
Onboard agentic filtering · Real-time SAR & optical · NVIDIA Jetson Orin · Under 15 W
14:57
UTC Time
15W
Power
24
Tiles
v0.1.0
◆ Triage Pipeline
◉ Live Feed
▤ Analytics
⛨ Audit Trail
◈ System Status
Tiles
24
Keep Rate
17%
Items
38
BW Saved
93.2%
Avg Power
8.4W
Latency
42ms
Defense ISR (Maritime) — Strait of Hormuz 8 TILES · S1 SAR
KEEP 3 FILTER 5 ★ 30 ITEMS BW · 91%
Tile 1: Vessel formation 27.37°N 55.06°E
KEEP
Score: 0.910
★ 2 ships, 1 dark-ship
Cloud
12%
Anomaly
74%
Value
89%
Step 1: HIGH urgency — 2 ships, 1 dark-ship
Step 2: KEEP full resolution + flag for downlink
Step 3: TRIGGER follow-up imaging on next pass
Tile 2: Unidentified vessel — no AIS
KEEP
Score: 0.987
⚠ 1 dark-ship
Cloud
8%
Anomaly
92%
Value
95%
Vessel ~80m detected, no AIS broadcast match in 3 h window.
⚠ Dark-ship anomaly — recommend cross-cue.
Tile 3: Open ocean — no targets
FILTER
Score: 0.122
Cloud
28%
Anomaly
5%
Value
12%
CNN-only decision — low value tile, no detections.
Action: COMPRESS to thumbnail. Bandwidth saved: 95%.
🔥 Wildfire Detection (California) — FIRMS overlay 8 TILES · S2 + FIRMS
KEEP 2 FILTER 6 ★ 5 ITEMS BW · 92%

Above is a CSS-rendered preview of the live dashboard. The actual app pulls real satellite data and runs the same triage pipeline in your browser.

Capabilities

Everything you need to ship

A complete onboard ISR stack — from sensor input to authenticated downlink decisions — with object detection, agentic reasoning, model versioning, and live satellite feeds built in.

Quantized CNN Inference

MobileNetV3-Small at INT8 via TensorRT > ONNX > PyTorch with auto-fallback. Outputs cloud, anomaly, and value scores per tile.

SAR Ship Detection

Classical CFAR + connected-components vessel detector for Sentinel-1 SAR. Identifies "dark ships" — vessels broadcasting no AIS transponder.

Agentic ReAct Reasoning

Pure-Python think → act → observe loop activated on high-value tiles only. Under 2 W on Jetson. Optional Phi-3/Gemma SLM enhancement on AGX/Thor.

HMAC Audit Trail

Every triage decision is logged with HMAC-SHA256 authentication and full provenance — tile hash, scene ID, scores, agent steps, timestamp.

Live Satellite Feeds

Pulls real Sentinel-1 SAR, Sentinel-2 L2A, NOAA GOES-18, and NASA FIRMS imagery. No paid API keys required for the public sources.

Model Registry & Rollback

SHA-256 checksums, training metadata, one-command rollback. Every deployed model is traceable to its training run and dataset.

Continuous Retraining

Ground-station feedback loop: analyst CSVs or audit-log pseudo-labels → fine-tune → validate → auto-register if improved.

Power Guard

Live tegrastats integration on Jetson. Agent loop only runs when value score > 0.6 AND power budget allows. Hard ceiling at 20 W.

Production CI/CD

GitHub Actions: lint → test → security scan → SBOM → Docker + Trivy. Multi-arch Dockerfile, hardened with non-root user.

Architecture

The data flow

From raw sensor input to authenticated downlink decision in under 50 ms.

1. Ingest

Tile-split, normalize
any spectral band layout

2. CNN

INT8 MobileNetV3
cloud · anomaly · value

3. Detect

YOLOv8 / SAR CFAR
items of interest

4. ReAct

Agentic reasoning
only on high-value

5. Decide

KEEP or FILTER
with explanation

6. Audit

HMAC-SHA256
JSON Lines log

Dual-Use Applications

Built for the missions that matter

The platform is sensor-agnostic and mission-aware. The same engine runs wildfire monitoring, maritime ISR, disaster response, and pipeline inspection.

Disaster · Climate

Wildfire Detection

Real-time hotspot and smoke-plume prioritization. Cross-references with NASA FIRMS active-fire feed for ground truth.

Defense · ISR

Maritime Surveillance

Sentinel-1 SAR ship detection over contested waters. Flags vessels with no AIS broadcast — the actual intel signal.

Humanitarian · HADR

Disaster Response

Post-event damage assessment. Filters out unchanged tiles, surfaces collapsed structures and refugee movement.

Commercial · Energy

Pipeline & Grid

Drone or satellite-based monitoring of pipelines, transmission lines, and offshore platforms for leaks and anomalies.

Developer Experience

Three lines to triage

The API is intentionally minimal. Drop in any image array; get back a structured decision with full provenance.

# pip install -e ".[ml,dashboard]" from edge_triage import EdgeTriageEngine engine = EdgeTriageEngine() result = engine.process_tile(image, {"context": "Maritime ISR"}) print(result.keep, result.final_score, result.detection_result.summary()) # True 0.91 "3 vessels (1 dark)"

Ready to see it in action?

The dashboard pulls real Sentinel-1 SAR over the Strait of Hormuz
and runs ship detection in your browser — clone & run in 60 seconds.

git clone && make install && make run — full source under MIT license.