AeroMesh

A decentralized, offline-first early warning system turning smartphones into edge-AI weather sensors and rescue beacons. It processes local data on-device and shares via P2P instead of satellites.

Comment

Inspiration

Natural disasters don't just destroy buildings; they destroy the invisible threads that keep us safe: our communication networks. When a cyclone or flood hits, the first things to go are the power grid and cellular towers. This creates a "Blackout Zone" where victims are silenced, and rescue teams are flying blind.

Growing up in regions where seasonal monsoons and severe storms are a reality, we realized that the most powerful tool for disaster response is already in everyone's pocket—a smartphone. We wanted to see if we could turn these isolated devices into a collective, intelligent, and decentralized safety web that works even when the world goes dark.

What it does

AeroMesh AI is an offline-first emergency response system that transforms smartphones into a distributed network of edge-AI weather sensors and rescue beacons.

  • Edge-AI Sensing: The app uses the phone's built-in barometer to monitor atmospheric pressure. A local TensorFlow Lite model analyzes these trends in real-time to detect impending severe weather anomalies without needing a cloud connection.
  • Decentralized Mesh Networking: If an anomaly is detected or a user triggers a manual SOS, the app uses Bluetooth Low Energy (BLE) to broadcast an encrypted, 18-byte "Heartbeat" payload. Nearby phones pick up this signal and "hop" it across the community, creating a viral alert web that extends far beyond the range of a single device.
  • The Gateway Bridge: If just one device in the mesh catches a faint cellular signal, it acts as a gateway, compressing the entire offline SOS queue and forwarding it via SMS to a central command dashboard for first responders.
  • Responder Heatmap: Rescue teams view a real-time Next.js dashboard that visualizes these mesh-relayed signals, providing an interactive heatmap of exactly where trapped individuals are located in the blackout zone.

How we built it

We architected AeroMesh AI to be resilient across five layers:

  • Mobile Client: Built with React Native (Expo Bare Workflow) and TypeScript for cross-platform hardware access.
  • Local Persistence: We used SQLite with Drizzle ORM to manage a time-series database of sensor readings and an SOS queue on the device.
  • Edge Intelligence: A custom TensorFlow Lite model was integrated for on-device anomaly detection.
  • Mesh Protocol: We developed a custom binary encoding format to fit GPS, Timestamps, and Alert types into the highly restrictive 31-byte BLE advertisement limit.
  • Command Center: The dashboard was built using Next.js 15, Tailwind CSS, and Mapbox GL JS, with a Twilio-powered SMS webhook to ingest data from the mesh gateways.

Challenges we ran into

  • The 18-Byte Constraint: Fitting precise GPS coordinates, a unique message ID, and a timestamp into a single BLE advertising packet was a massive hurdle. we had to utilize bit-packing and coordinate quantization to squeeze maximum data into every byte.
  • Background Resilience: Modern mobile operating systems are aggressive about killing background tasks to save battery. We had to implement complex BackgroundFetch and Task Manager logic to ensure the "Pulse" stays alive without draining the user's phone during a disaster.
  • Cross-Platform Parity: Handling the differences between Android and iOS Bluetooth stacks—specifically background peripheral advertising—required deep native bridging and iterative testing.

Accomplishments that we're proud of

  • True Offline Communication: Seeing an SOS signal "hop" from a phone in airplane mode to another phone, and eventually to the dashboard, felt like magic.
  • Graceful Degradation: We successfully built a system that works on flagship devices with high-end sensors but still provides life-saving relay utility for budget devices without barometers.
  • Binary Efficiency: Achieving a 40% reduction in payload size using MessagePack encoding for the SMS gateway bridge.

What we learned

  • Decentralized Thinking: We learned to stop relying on "The Cloud" as a given. Designing for a world where the internet doesn't exist forced us to be much more efficient with data and power.
  • Sensor Fusion: We gained deep insights into how environmental sensors (barometers, light sensors, accelerometers) can be combined to create a high-confidence picture of a user's physical reality.
  • Human-Centric UI: We learned that in a crisis, less is more. Designing an OLED Dark Mode interface taught us how to balance battery preservation with psychological comfort.

What's next for AeroMesh

  • Multi-Modal Fusion: Integrating the microphone and accelerometer to detect earthquakes and structural collapses automatically.
  • LoRA Integration: Bridging the smartphone mesh with LoRA (Long Range) hardware to extend the mesh range to several kilometers.
  • Government Partnerships: Collaborating with disaster management agencies to standardize the AeroMesh protocol, making every citizen a part of the national early-warning infrastructure.
Share this project:

Updates