Neighbor Needs — Devpost Submission

Inspiration

When a wildfire, flood, or earthquake hits, the first 72 hours are the most critical — and official emergency services are almost always overwhelmed. We kept asking ourselves: what if your neighbor has a generator, medical training, or a spare room, but you just don't know? That question became Neighbor Needs. We wanted to make the invisible safety net of a neighborhood visible, without compromising the privacy that makes people willing to share in the first place.

What it does

Neighbor Needs is a neighborhood disaster resilience skills map. Residents register their:

Skills — medical, construction, mechanical, language interpretation, etc.
Equipment — generators, chainsaws, first aid kits, water purification
Shelter capacity — spare rooms, accessible spaces

In normal mode, neighbor locations are anonymized and contact info is hidden. When an emergency is declared, the map activates within an affected radius: pins turn red, contact information becomes visible to nearby neighbors, and help requests begin flowing — turning a passive directory into a real-time mutual aid network.

How we built it

Frontend

We built a React + TypeScript frontend with TailwindCSS v4 for styling and a set of shadcn-style UI primitives backed by Radix UI. The map uses CSS-based lat/lng positioning to simulate neighbor proximity:

offsetX = (lng − center.lng) × 400
offsetY = (lat − center.lat) × 400

Backend

We built a Go backend using Gin for routing and pgx/v5 for PostgreSQL connections. Authentication uses JWT access tokens (15min TTL) with device-bound refresh token rotation, and passwords are hashed with bcrypt. The database uses PostGIS for geospatial queries-ST_DWithin powers the proximity search that finds neighbors within a configurable radius. Configuration is managed via Viper from a YAML file, and structured logging uses Zap. The service layer pattern collapses handlers and business logic into single-file services for auth, user profiles, discovery, and emergency coordination.

Mobile Apps

We built a SvelteKit + TypeScript mobile app with TailwindCSS v4 and shadcn-svelte components, wrapped in a Tauri 2 shell for cross-platform native deployment. The map uses Leaflet with OpenStreetMap tiles for neighbor visualization, and Lucide icons provide UI iconography. The static adapter builds a self-contained bundle that Tauri serves from its local webview, enabling offline-first operation. Tauri commands expose native system capabilities like desktop notifications for emergency alerts, while the Rust backend handles platform-specific integrations. The mobile prototype mirrors the web frontend's data model with mocked responses in the Svelte pages.

Challenges we ran into

The hardest problem was the privacy-utility tradeoff. Showing too much by default erodes the trust needed for adoption. Showing too little during an emergency costs lives. We landed on a two-mode system where privacy escalation — revealing phone numbers and precise coordinates — is:

Gated behind an explicit emergency declaration
Scoped to a geographic radius, not broadcast globally
Designed to move fully server-side in production, so the frontend never receives data it shouldn't render

We also had to think carefully about abuse vectors: a fake emergency declaration, or a bad actor registering false skills. These informed our "What's next" roadmap around verified registration.

Accomplishments that we're proud of

The emergency mode UX. A single toggle transforms the entire interface — banner color, pin states, sidebar content, and contact visibility — so the shift from normal to crisis feels immediate and unambiguous. We're also proud of designing a privacy model that doesn't treat safety and discretion as opposites. End-to-end encryption for communication between the client and server was also a big accomplishment.

What we learned

Hyperlocal tools need to earn trust before they need to earn adoption.

The constraints we almost removed — anonymizing by default, scoping visibility by radius — turned out to be the most important features. We also learned that geospatial privacy is genuinely hard: even anonymized pin clusters can reveal home locations if the anonymization radius is too small.

What's next for Neighbor Needs

Verified registration via SMS or community vouching to prevent misuse
Server-side privacy escalation so precise location data never leaves the backend until an emergency is active
Offline-first mode via the Tauri shell, so the app works when cell networks are congested
Pilot program with a real neighborhood association or CERT team