HALO | Devpost

Inspiration

After Hurricane Katrina, the collapse of community support networks drove an 89% rise in serious mental illness and a 62% rise in suicidality in New Orleans. After Hurricane Ida, 36.5% of food distribution sites operated for just one day — not because help wasn't there, but because no one knew where to find it.

We were struck by a painful irony: during climate disasters, the problem is rarely a shortage of goodwill. Neighbours have spare insulin. Restaurants have leftover food. Volunteers are ready. But without a way to surface that local capacity in real time, it stays invisible. We built HALO to fix that.

What it does

HALO is an AI-powered, community-driven platform that transforms fragmented disaster information into actionable support. It provides:

A live map of five critical resources during climate emergencies: shelter, clean water, food, medical care, and mental health support across New Orleans
Real-time community reporting so residents and responders can post supplies available, road blockages, drop-off points, and needs as they emerge
AI-powered routing that avoids community-reported blockages and directs users to the most relevant nearby resource
A sustainability layer where people with surplus resources (restaurant food, medications, baby supplies) can post drop-off points for redistribution
An AI chatbot powered by Claude that interprets free-text needs and connects users to help in plain language
Full accessibility including voice input, text-to-speech, colour blind mode, high contrast, and simple language mode, because the most vulnerable are also most likely to face digital barriers

When a user searches for medication, results come from community reports first, because during a flood, a neighbour with spare insulin is more useful than a pharmacy that's underwater.

How we built it

Frontend: React via Lovable, mobile-first and fully responsive
Map: Leaflet.js with OpenStreetMap tiles, centred on New Orleans
AI chatbot: Claude API for natural language resource matching and signposting
Real-time database: Supabase for community reports with live updates
Real data sources: - OpenFEMA API: disaster declarations and official shelter locations for Louisiana - New Orleans ArcGIS Open Data Portal: FEMA flood zone maps (FIRM), community facilities, health clinics - data.nola.gov: essential services and drug store locations - NOLA Ready: live emergency alerts for New Orleans - OpenStreetMap: base map and routing

Challenges we ran into

Data fragmentation The problem we're solving is also the problem we faced building. Official data for New Orleans is spread across FEMA, ArcGIS, data.nola.gov, and NOLA Ready with inconsistent formats, requiring significant normalisation work
Trust and verification Community reporting is only useful if people trust it. Designing a lightweight verification system (confirmation scores, reporter proximity weighting, human-in-the-loop flagging) without adding friction for users in crisis was a hard design problem
Routing around dynamic blockages Making routes respond to real-time community hazard reports required custom integration with Leaflet Routing Machine beyond standard usage
Accessibility under pressure Designing for users who may be panicked, elderly, or have low digital literacy meant every interaction had to be ruthlessly simplified without losing functionality
Scope We had a lot of ideas and 48 hours. Deciding what to cut was as hard as deciding what to build

Accomplishments that we're proud of

Built a working, real-data product in 48 hours as a team of five who had never worked together before
Successfully integrated live government APIs (OpenFEMA, NOLA ArcGIS, data.nola.gov) rather than relying on dummy data
Designed a platform that works for both sides of a disaster: people who need help and people who want to give it
The sustainability redistribution layer, turning surplus restaurant food and spare medications into community assets rather than waste, emerged organically from our discussions and ended up being one of our favourite features
Built full accessibility into the core product from day one, not as an afterthought

What we learned

Community resilience is not the absence of disaster. It's the speed at which connections reform after one hits
The hardest UX challenge isn't making something powerful, it's making something usable by someone who is scared and has 30 seconds
Real open data is messier than it looks but far more compelling in a pitch than anything fabricated
A team with diverse skills (CS, business, psychology) makes better product decisions than a homogeneous one. Our psychology background shaped the mental health layer and our business background kept the feasibility story honest

What's next for HALO

Global expansion HALO's open data model means any city with public geospatial and emergency data can deploy it. Next targets: cities with similar vulnerability profiles to New Orleans including Dhaka, Lagos, and Manila
UN International Charter on Space and Major Disasters Integrate satellite imagery partnerships for real-time flood and damage mapping
AI verification Move from community confirmation scores to AI-enabled content moderation with geofence checks and human-in-the-loop review at scale
White-label deployments for municipal emergency management offices
Sustainable funding through national resilience grants and CSR partnerships with logistics, insurance, and telecoms firms, sectors with direct financial incentive to reduce disaster impact
Ultimately: making HALO the infrastructure layer that every community-powered disaster response runs on