Sentinel-g

🛰️ About Sentinel-G 💡 Inspiration

Northeast India faces frequent floods and landslides, where delayed or inaccurate information often costs lives. During such disasters, responders struggle to establish ground truth due to fragmented reports, poor connectivity, and delayed satellite insights. Sentinel-G was inspired by this gap — the need for a single, reliable, real-time intelligence layer that works even when infrastructure fails.

🛠️ How We Built It

Sentinel-G is a multimodal, offline-first crisis command center that fuses satellite data, AI reasoning, and community inputs.

We used Google Gemini 3.0 (Flash & Pro) for multimodal analysis, enabling the system to interpret images and audio to identify disaster type, severity, and approximate impact radius. A React + TypeScript dashboard visualizes this intelligence using Leaflet heatmaps, historical timelines, and live weather overlays.

For emergency response, we integrated:

SOS beacons that share live GPS location and battery status via WhatsApp/SMS

Crowd-sourced alert verification to filter misinformation

Evacuation routing using OSRM

Resource mapping via OpenStreetMap (Overpass API)

The application is built as a Progressive Web App (PWA) to ensure usability in low-connectivity disaster zones.

📚 What We Learned

Designing AI systems for high-stakes environments requires prioritizing reliability over model complexity.

Multimodal reasoning significantly improves situational awareness compared to single-input systems.

Offline-first architecture is essential for real-world disaster response, not an optional feature.

Clear visualizations and simple interactions matter more than dense technical detail in emergencies.

🚧 Challenges We Faced

Synchronizing real-time crowd verification without introducing alert inconsistencies.

Balancing AI inference speed vs. accuracy during live disaster analysis.

Designing a UI that remains usable under stress, low bandwidth, and partial data availability.

Integrating multiple open data sources with differing update rates and reliability.

🌍 Impact

Sentinel-G demonstrates how AI can transform fragmented disaster data into actionable intelligence, helping authorities and citizens respond faster, safer, and with greater confidence when every minute matters.