TerraSense
Built at AI ATL 2025 Hackathon
Geospatial intelligence platform that processes NASA satellite data to analyze vegetation health. Users draw custom regions on an interactive 3D globe to access 25 years of NDVI metrics and AI-powered predictions.
Architecture
Frontend: React.js + CesiumJS for 3D globe visualization and region selection
Backend: FastAPI handling satellite data processing and API orchestration
Data: Google Earth Engine API for NASA satellite data, MongoDB Atlas for storage
AI: Google Gemini API for trend analysis and vegetation predictions
Technical Approach
Implemented intelligent batching and interpolation to handle thousands of satellite data points per region while managing Google Earth Engine API rate limits. Designed prompt engineering pipeline for Gemini to generate evidence-based predictions from historical NDVI patterns.
Team
Built with Kevin Kou and edwinthedev at AI ATL 2025.
Links: Devpost