💡

Inspiration

During natural disasters, two critical problems arise: cellular network infrastructure frequently collapses, and search-and-rescue teams are overwhelmed with distress signals without knowing who needs help most urgently. We built SafeSearch to bridge this gap, ensuring that victims can broadcast their location and medical status even without internet, while rescue teams can instantly triage and prioritize operations based on real-time urgency and data. 🚀

What it does

SafeSearch is a full-stack, AI-prioritized disaster response and victim locator system featuring: AI Triage & Prioritization: A fast triage engine categorizes incoming messages based on victim count, injuries, and entrapment status, ranking them dynamically (Critical, High, Medium, Low) to ensure rescuers help the most vulnerable first. Offline SOS Mode: When cell towers are down, the system shifts to a zero-network protocol. It compiles vital medical info (blood group, conditions) and coordinates into a high-density, easily scannable QR code for rescuers or drones, or prompts the user to send a formatted SMS. Live Interactive Map: Plots real-time victim coordinates alongside live global satellite feeds (wildfires, storms, earthquakes, volcanoes) fetched directly from the NASA EONET API. Real-Time Tracking & Live Feed: Uses WebSockets to connect victims and rescuers. Victims get real-time tracking of their assigned rescue team’s distance and ETA. Rescuer & Admin Dashboards: Command dashboards for rescuers to navigate, play on-scene video feeds, and manage rescue statuses, coupled with a central analytics center for government administrators. 🛠️

How we built it

We engineered SafeSearch as a full-stack distributed system: Frontend: Built using React, Vite, and styled with TailwindCSS. Maps are rendered using React-Leaflet and OpenStreetMap. Backend API: Powered by Node.js & Express handling user authentication and data management. Real-time, bi-directional messaging is powered by Socket.io. Database: MongoDB with Mongoose schemas stores user profiles, SOS alerts, and rescue task logs. AI Triage Service: Built with FastAPI (Python), utilizing regex-based text analysis and keyword parsing to calculate severity scores and generate context-aware first-aid instructions. 🚧

Challenges we ran into

Robust Geolocation: GPS can fail or be blocked. We solved this by creating a multi-layer fallback: trying browser GPS first, falling back to IP-based location services, and finally letting the user manually input landmarks or coordinates. Network-less SOS Transmission: Developing a way to package and send data with zero internet. We resolved this by compressing vital payload data into structured JSON and embedding it into high-density dynamic QR codes. 🎉

Accomplishments we're proud of

Designing a complete offline-to-online safety loop using QR codes and SMS redirects. Integrating live satellite feeds from NASA EONET in real time. Creating a clean, modern user interface that feels premium, highly readable, and functional under stress. 🧠

What we learned

How to design applications for high-reliability under extreme environmental constraints. Managing client-side offline states dynamically in React. 🔮

What's next

Edge LLMs: Migrating the FastAPI keyword parser to a localized, lightweight LLM running directly on the victim's device. LoRaWAN Integration: Allowing SOS data transmission over long-range, low-power radio networks. Drone Scanning: Integrating computer-vision drone routing to automatically fly over affected areas and scan QR codes displayed on victim screens from the air.