A real-time dashboard that scores prescribed-burn readiness from live sensors and satellite data.
- React
- FastAPI
- scikit-learn
- ESP32
- NOAA
- Sentinel NDVI
- Supabase
THE PROBLEM
Prescribed burns keep wildfire risk down, but deciding when a burn is safe means juggling weather, soil moisture, fuel dryness, and satellite data by hand. Get the window wrong and a controlled burn becomes a real fire.
THE BUILD
I built a real-time dashboard that scores burn readiness and predicts safe burn windows. It pulls live readings from an ESP32 sensor rig I wired up (temperature, humidity, soil moisture), then combines that with NOAA forecasts, Sentinel Hub NDVI imagery, and NASA FIRMS fire data. A scikit-learn model turns all of it into a go or no-go call, served through a FastAPI backend with Supabase behind it.
THE RESULT
PyroLens earned 2nd place in the Assurant Challenge and 3rd in Capgemini Net Positive AI at the KSU Hackathon for Social Good. The dashboard and the API are deployed and live.