Belun NGO · Conflict early-warning · Civil society
Belun AtReS
Timor-Leste's most complete conflict-analysis platform — a Postgres-backed graph of 7,200+ incidents, 98,500+ relationships, with AI Tetun-English translation and full municipal drill-down. Built and operated for Belun, the country's leading conflict-prevention NGO.

The problem we solved
Belun runs a national early-warning system for community-level conflict. Field staff submit incident reports from across all 13 municipalities, in Tetun. The reports need to be coded, cross-referenced, and aggregated so policy users — UN, embassies, donor governments, the National Police, civil society partners — can see patterns at municipal, sub-post, and actor level. The previous system was a flat spreadsheet. The international audience couldn't read the source data.
What we shipped
Connected, not flat
Incidents, people, organisations, weapons, and locations are linked together, so a question like "where has this group been involved before?" is two clicks, not an afternoon of cross-referencing.
Tetun reports, English insights
Field reports come in from Belun staff in Tetun. The system translates them to English automatically using our own translation engine, so international policy users read the same data.
Municipal drill-down
Every one of the 13 municipalities and 65 administrative posts has its own view — incident frequency, type, severity, trend over time, and the patterns of weapons involved.
Built for real use, not demos
Role-based access, audit trails, encrypted sensitive fields, controlled exports, nightly backups. The data is sensitive and the access patterns reflect that.
How it works, end to end
Belun's field staff submit incident reports through a structured form. Each report is automatically translated from Tetun to English on the way in — both versions are kept, so the original is never lost. A coding team reviews each report for quality, deduplication, and consistent classification. The public-facing dashboard then reads the cleaned data and presents it as maps, trends, and municipal-level breakdowns that policy users can navigate without ever needing to know the underlying shape of the data.
Why it lasted
Belun AtReS has been operating for years and remains the authoritative reference for community-level conflict data in Timor-Leste. It survives because the operating model is boring on purpose — the same hosting, the same nightly backups, the same upgrade rhythm as every other OniT-operated platform. Boring infrastructure is what lets the data on top of it stay interesting.
Stack