AegisHimal

Inspiration

Living near and traveling through the Himalayan corridors—such as the Dehradun–Rishikesh–Tehri highway network—reveals a stark vulnerability: during monsoons and cloudbursts, severe landslides and flash floods occur precisely when cellular networks collapse. Power transmission poles topple, base stations lose backhaul fiber, and travelers are left stranded in a communication blackout.

In these critical first hours, conventional emergency platforms fail because they make a flawed assumption: constant, high-bandwidth cloud connectivity. Without signal, motorists drive directly into washed-out roads, and isolated mountain communities cannot broadcast their status.

We asked ourselves: Why should disaster response and road-hazard alerting grind to a halt just because cell towers lose power?

We were inspired to build AegisHimal—an open-source, local-first progressive web application (PWA) that converts commodity smartphones into autonomous, decentralized hazard sentinels capable of operating, verifying, and propagating emergency alerts peer-to-peer with unstable internet connection.

What It Does

AegisHimal enables offline reporting, edge-based hazard validation, and delay-tolerant data dispatch:

1. Zero-Connectivity Hazard Reporting: Users log road washouts, rockfalls, or bridge collapses offline. A built-in edge vision model validates images on-device to filter out non-hazard photos.
2. Opportunistic Peer-to-Peer Mesh Sync: Reports are packaged into lightweight, cryptographically signed bundles. Devices sync peer-to-peer with passing vehicles and hikers via local WebRTC and animated high-density QR handshakes.
3. Mule-to-Cloud Ingestion: As soon as any participating device enters a cellular or Wi-Fi pocket, it automatically dumps its synchronized queue into the central ledger.
4. 3D Topographic Vulnerability Engine: A browser-based GIS dashboard renders terrain elevation, dynamically computing slope gradients and landslide susceptibility indices across mountain transit corridors.

Built With

• dexie.js
• go
• maplibre-gl-js
• onnx-runtime-web
• postgresql
• react
• typescript
• webrtc
• workbox
• yjs