CloudAlert Bridge

In Bangladesh, sudden-onset disasters like catastrophic fires, flash floods, and severe weather events pose a continuous threat to millions. During these critical emergencies, timely communication is the difference between life and death. However, official government alerts and news updates are often published online or broadcasted broadly, failing to immediately reach the specific individuals in the direct path of danger. Who is affected: The most vulnerable populations—particularly those in densely populated urban areas or remote coastal regions—often lack reliable internet access or smartphones. They rely on basic mobile phones, rendering modern web-based early warning systems ineffective for them. When it matters: The technological disconnection between digital hazard data and offline vulnerable communities is most fatal in the immediate "golden hour" of a rapidly unfolding emergency, when hyper-localized, actionable evacuation information must be disseminated instantly. Currently, there is a critical disconnect between the generation of digital hazard alerts and the offline delivery of those warnings to the precise geographic locations at risk.

What We Built: CloudAlert Bridge is an early warning system designed for Bangladesh that sends life-saving SMS alerts to communities during natural disasters. Built on a high-speed Python (FastAPI) and spatial database (PostGIS) backend, it translates complex hazard data into immediate, targeted text messages in Bangla, bridging the gap between digital disaster tracking and offline populations. How It Works The platform operates through a four-stage pipeline designed for speed, accuracy, and reliability: 1. Dual Data Ingestion (Gathering Hazard Data) The system collects emergency information in two ways: -> Automated Scans: Background processes constantly monitor official tracking sources—including the Bangladesh Meteorological Department (BMD), the Flood Forecasting and Warning Centre (FFWC), and the USGS—to catch official weather and disaster alerts instantly. -> Manual Entry & Community Reports: Human operators can input specific emergency events directly. Additionally, crowdsourced "witness reports" can be submitted, but these go through a strict moderation queue and require administrator approval before moving forward. 2. Geospatial Targeting & Safety Controls (Finding the Danger Zone) Once a hazard is active, PostGIS (a spatial database tool) calculates the exact geographic area affected. -> Tower Matching: The system matches the danger zone against mobile network coverage maps to pinpoint which cell towers cover the area. -> Two-Person Approval: For manually created severe threats, the system requires approval from two separate administrators to prevent false alarms and public panic. 3. Mass Simulation & Telemetry (Testing Network Load) Because sending millions of texts requires official partnerships with national telecom providers (such as Grameenphone, Robi, and Banglalink), the platform includes a built-in simulation engine. -> It models the stress of a mass text broadcast on cellular networks. -> It tracks projected delivery speeds and provides detailed logs on any simulated broadcast failures. 4. Live SMS Demonstration (Real-World Testing) To prove the system works in reality rather than just theory, the backend connects to the sms.net.bd API. -> By whitelisting specific real phone numbers, the platform bypasses the simulation layer for those devices. -> It sends actual, physical SMS alerts directly to real mobile phones, proving its capability for real-world emergency communication.

CloudAlert Bridge is built using a modern, asynchronous Python stack designed for high throughput and complex geospatial calculations: #Core Frameworks: -> FastAPI: The core backend framework, chosen for its asynchronous capabilities and speed, handling both the public APIs and the dispatcher console. -> Uvicorn: The lightning-fast ASGI web server running the FastAPI application. -> Python 3.11+: The foundational programming language. #Database & Geospatial: -> PostgreSQL with PostGIS: Used as the primary database, leveraging PostGIS for advanced geospatial querying (e.g., intersecting hazard zones with cell tower coverage areas). -> Shapely: Used for complex Python-side geometric operations, such as calculating toxic plume trajectories and hazard radiuses. #Data Ingestion & Scraping: -> BeautifulSoup4 & Feedparser: Used by the crawler to parse live hazard feeds (like BMD CAP feeds) and extract unstructured fire incident data from the web. -> HTTPX & Requests: For making asynchronous and synchronous HTTP requests to external portals and APIs. #Third-Party Services & APIs -> sms.net.bd: Integrated for the live, real-world demonstration of SMS delivery in Bangladesh, bypassing the system's internal simulator to prove real-world viability.

The architecture, development, and debugging of the CloudAlert Bridge were significantly accelerated using advanced Large Language Models. Special thanks to: -> Google Gemini: Provided extensive assistance in writing core backend logic, architecting the dual-ingestion pipeline, and debugging geospatial routing. -> Anthropic Claude: Assisted in structuring the project, formulating the problem statements, and refining the system design constraints.