Humanitarian · UNDP
RAPIDA: Community Crisis Damage Assessment
Client
United Nations Development Programme
Location
Global
Year
2024
Services
Custom Development
For UNDP's RAPIDA early recovery methodology, Kaistrum engineered an end-to-end crowdsourced crisis-response platform — enabling communities to submit geolocated damage reports within 72 hours of a sudden-onset disaster, with AI-assisted translation across all six official UN languages and offline-first mobile functionality for low-connectivity environments.
RAPIDA crisis damage map — geolocated community submissions
500k+
Report capacity per crisis
Backend load-tested to 500,000 concurrent submissions — matching UNDP's large-scale crisis requirement.
6
UN languages
All six official UN languages supported via AI-assisted translation — no manual localisation required.
72hr
Deployment readiness
Platform can be activated for a new crisis in under 72 hours with no per-crisis configuration.
100%
Offline capable
Submissions cached locally and synced when connectivity returns — functional with zero internet.
The Challenge
72 hours to understand a disaster. No tool existed.
UNDP's RAPIDA methodology required community-sourced damage data within 72 hours of a crisis event to guide early recovery resource allocation. No single tool existed that was simultaneously multilingual, offline-capable, scalable to 500,000+ reports per crisis, and exportable in UNDP's standard interoperable formats — and deployable across hundreds of crises per year without per-crisis configuration.
- 0172-hour assessment window — but field survey tools required days to configure and deploy
- 02Communities span Arabic, Chinese, French, Russian, Spanish, and English — no single tool supported all six
- 03Low and zero connectivity environments in the most affected areas
- 04Existing tools broke at 50,000 submissions — UNDP needed 500,000+ per crisis, 100s of crises per year
- 05Outputs needed to be in GeoJSON, CSV, and shapefile formats for UNDP geospatial systems
The Solution
End-to-end crowdsourced crisis assessment.
Kaistrum built an open-source, end-to-end platform: a community-facing app accepting photo, GPS, and damage classification submissions via smartphone or WhatsApp; an AI-assisted translation and image classification pipeline; and a backend dashboard with live map and structured data export.
01
Community submission app
Web and WhatsApp-based submission flow accepting photos, GPS coordinates, damage classification, and infrastructure type — in all 6 UN languages with AI-assisted translation of free-text descriptions.
02
Building footprint map overlay
Interactive map with building footprints from OpenStreetMap and UNDP datasets — allowing submitters to geolocate damage to specific structures without needing precise GPS.
03
AI image classification
Computer vision pipeline classifying submitted photos as minimal, partial, or complete damage — reducing analyst review burden and enabling automated batch processing.
04
Scalable backend
PostgreSQL/PostGIS backend with horizontal scaling — load-tested to 500,000 concurrent report submissions. Structured for reuse across hundreds of crisis deployments per year.
05
Analytics dashboard
Live GIS dashboard displaying submissions with geolocation, damage level, and infrastructure type — with one-click export to GeoJSON, CSV, and shapefile.
Damage assessment mobile submission flow
Outcomes
Report capacity per crisis
Backend load-tested to 500,000 concurrent submissions — matching UNDP's large-scale crisis requirement.
UN languages
All six official UN languages supported via AI-assisted translation — no manual localisation required.
Deployment readiness
Platform can be activated for a new crisis in under 72 hours with no per-crisis configuration.
Offline capable
Submissions cached locally and synced when connectivity returns — functional with zero internet.
Work with Kaistrum