Crisis Connect

Inspiration

Today, over 2.3 billion people face food insecurity, and 123 million people are displaced by conflict and disaster. During crises, people often struggle to find help quickly and safely.

This inspired us to create Crisis Connect — a map-first platform that connects people who need help with those who can provide it in real time.

Existing platforms such as Ushahidi and Nextdoor either focus on documenting crises or only support small local communities. However, there is still no platform that can coordinate help at scale while protecting users' privacy. Crisis Connect aims to bridge that gap by enabling people to request or offer help on a live map while keeping their identities and precise locations private.

What It Does

Crisis Connect enables people to request or offer help while keeping their identities and precise locations private.

Users can post requests or offers directly on a live map, where each post appears as a colored dot based on its urgency
AI analyzes each post and assigns a priority score to highlight the most critical situations
Users can search for nearby needs using natural language
In-app chat lets people communicate without sharing phone numbers or emails
Voice features allow people to submit requests by speaking and listen to AI-generated crisis briefings, making the platform more accessible during emergencies when typing or reading may be difficult

How We Built It

Crisis Connect was built in 24 hours using a modern full-stack architecture.

Frontend — React, TypeScript, Tailwind CSS, and Leaflet for a map-first interface
Backend — NestJS and MongoDB on MongoDB Atlas, using geospatial queries to fetch nearby posts
Real-time — Socket.IO for live updates and chat
AI — Claude for urgency prioritization
Voice — ElevenLabs for speech-to-text posting and audio crisis briefings
Deployment — Vercel and Railway

Challenges We Ran Into

Complexity at speed — coordinating the map interface, backend APIs, and real-time updates required careful integration to ensure posts appeared instantly for all users
Geospatial queries — this was our first time building a map-based application; implementing proximity-based lookups was new territory for our team
Reliable AI outputs — AI responses are not always structured, so making urgency scoring and search filters consistent required extra work
Privacy vs. utility — balancing useful location information with user privacy, ensuring the map shows approximate areas without revealing exact coordinates