AI Verification & Analysis Engine
The "Agentic Workflow" — a 4-stage automated pipeline that processes every news event from initial signal detection through satellite verification to confidence scoring.
Signal Triangulation
Ingestion Layer
Multi-source data ingestion aggregates conflict, news, and OSINT signals from diverse sources across the MENA/GCC region.
ACLED API
Structured event data on political violence and protests across Africa, Middle East, and Asia.
Liveuamap API
Real-time conflict mapping with geolocated events and territorial control data.
GDELT Project
Global Database of Events, Language, and Tone — monitoring broadcast, print, and web news.
Telegram OSINT
Official military and regional channels with custom scraping and translation logic.
Parseek
Arabic and Persian media aggregator for regional news coverage not available in English sources.
Geospatial Auto-Tagging
NER Processing
Named Entity Recognition (NER) extracts precise coordinates from unstructured text, automatically snapping news icons to pre-defined infrastructure points.
NER Engine
Custom-trained multilingual NER model optimized for Arabic, Persian, and English geographic entities.
Critical Infrastructure DB
Pre-defined GPS coordinates for refineries, ports, desalination plants, data centers, and military installations.
Snap Logic
Proximity-based algorithm that matches extracted locations to nearest critical infrastructure within configurable radius.
Satellite Ground Truth
Verification Layer
Multi-modal satellite imagery provides all-weather, 24/7 structural change detection and damage assessment for ground truth verification.
Sentinel-2 (Optical)
Copernicus Data Space Ecosystem — fire/damage detection through spectral analysis of optical imagery.
Sentinel-1 (SAR)
Synthetic Aperture Radar for 24/7 structural change detection through cloud, smoke, and darkness.
Planet/SkyWatch
0.5m resolution imagery for detailed verification of specific "Hot Zone" events and infrastructure damage.
Confidence Scoring & Summarization
AI Analysis
LLM-powered analysis assigns confidence scores based on the full evidence chain, generating multilingual geopolitical summaries.
LLM Engine
Large language model for evidence analysis, confidence scoring (Verified/Unverified/Propaganda), and geopolitical context.
Evidence Drawer
Complete chain of evidence from all pipeline stages, forming the "Digital Chain of Custody" for each event.
Multilingual NLP
English, Arabic, and Farsi summary generation with cultural context preservation and regional terminology.
Confidence Score Distribution
Corroborated by 2+ independent sources AND satellite imagery confirms ground truth.
Single-source report OR awaiting satellite confirmation. Displayed with caution indicator.
Source analysis and cross-referencing indicate deliberate misinformation or state propaganda.