GeoIntel

AI-driven satellite analytics for agriculture. GeoIntel monitors crop health, predicts yields, and provides early drought warnings to ensure food security.

Comment

Inspiration

Climate change and unpredictable weather patterns are making agriculture increasingly difficult, threatening food security and farmers' livelihoods. We realized there was a critical need for accessible, data-driven tools that could provide early warnings for droughts and accurate yield forecasts. Our goal was to empower governments, agricultural funds, and farmers with actionable intelligence drawn from space.

What it does

GeoIntel is an advanced agricultural intelligence platform. It ingests and processes massive amounts of satellite imagery and meteorological data from NASA MODIS, ESA Sentinel-2, and CHIRPS. By leveraging these world-class datasets, we compute critical vegetation indices like VCI (Vegetation Condition Index), TCI (Temperature Condition Index), and VHI (Vegetation Health Index). The platform provides a comprehensive dashboard to monitor crop health across different administrative units, track historical extremes, issue automated drought alerts, and generate accurate yield forecasts to prevent agricultural crises before they happen.

How we built it

We built GeoIntel with a robust, scalable architecture. The backend is a modern Python API using FastAPI, deployed on Google Cloud Run for serverless scaling. We leverage Google Earth Engine (GEE) as our primary data processing engine to rapidly analyze geospatial data and compute complex masks (e.g., cropland areas). Our data is stored in a highly available Google Cloud SQL PostgreSQL database with PostGIS for spatial queries. We also implemented a Cloud Scheduler batch pipeline that automatically triggers daily data ingestion and metric calculations.

Challenges we ran into

Processing historical satellite data to establish accurate baseline extremes for the VHI index required optimizing spatial queries and managing large datasets without hitting API rate limits.

Accomplishments that we're proud of

We successfully built a fully automated pipeline that goes from raw satellite imagery in Google Earth Engine to actionable business intelligence in a production database, entirely hosted on Google Cloud Platform. We're incredibly proud of our high-accuracy cropland mask generation and our automated scheduled batch jobs that ensure the data is always up-to-date without manual intervention.

What we learned

We learned how to orchestrate complex spatial workloads using Google Earth Engine within a microservices architecture. We also gained deep experience with GCP's serverless ecosystem (Cloud Run, Cloud Scheduler, Cloud SQL) and how to manage secrets securely in production environments.

What's next for GeoIntel

We plan to integrate more sophisticated machine learning models for yield prediction based on historical VHI trends. We also aim to expand our monitoring capabilities to track soil moisture (SMAP) and incorporate hyper-local weather forecasts to provide even more precise advisory alerts directly to farmers.

Built With

Share this project:

Updates