Hackathon · 2026 · Unicorn Mafia × Techbible Hack Night · London 2026
WILDSCAN
AI agents that scan online marketplaces for wildlife trafficking, decoding the coded language sellers hide behind in eight languages.
Overall Winner - Unicorn Mafia x Techbible Hack Night 2026
Built in 90 minutes; awarded the overall prize at the London hack night (Web MCP Agents track), March 2026.
Python · LangGraph · Next.js · PostGIS
The problem
Wildlife traffickers sell openly on regional marketplaces, hiding behind coded words in dozens of languages. Human investigators cannot keep up with the volume.
What I built
I built a pipeline of six agents that scrapes listings, spots coded terms from a 500 word lexicon in eight languages, identifies the species from photos, and scores every listing against records of real seizures.
What it does
- Six Agents, One Hunt. A scanner, a triage agent, a linguist, an image analyst, a species classifier, and a risk scorer run the whole pipeline on their own, built with LangGraph.
- Speaks the Sellers' Code. It knows 500 coded trafficking terms across eight languages, and still catches them when sellers misspell them on purpose.
- Scores You Can Check. Every listing gets a 0-100 risk score from eight weighted signals, checked against 6,000 records of real seizures.
- A Live Globe. Every detection lands on a Mapbox globe in real time, with an intelligence brief one click away.
Gallery
The full engineering is on GitHub.