Hackathon · 2026 · Claude Hackathon · Imperial College London 2026
DarkFleet
AI-powered maritime surveillance system that consolidates seven data sources to detect illegal fishing vessels via deterministic risk scoring and Claude-generated intelligence briefs.
6th of 64 - Claude Hackathon, Imperial College London
Top-10 finish out of 64 teams at the Anthropic-sponsored hackathon at Imperial College London.
Python · Claude API · PostGIS · Mapbox
Challenge
Illegal fishing fleets hide across fragmented signal sources - AIS dark gaps, RFMO records, sanctions lists, ownership chains. No analyst can correlate them in real time.
Approach
Unified seven open data sources (Global Fishing Watch, WDPA, RFMO registries, OpenSanctions, and more) under a single risk dashboard with 7-signal weighted scoring and a Claude-driven analyst persona for follow-up investigation.
What it does
- 7-Signal Risk Engine. Encounters, AIS dark gaps, RFMO status, loitering, flag changes, ownership opacity, and sanctions matches.
- Tiered Alerts. Red ≥80, Amber ≥60, Clear <60 - ranked across the global fleet in real time.
- Claude as Analyst. Streaming chat persona answers analyst follow-ups and produces structured vessel assessments.
- Incident Reports. Auto-generated PDF briefs synthesising every signal into a single shareable artifact.
Gallery
Photos
