Abstract:
This project, led by Earth Lab at the University of Colorado Boulder in partnership with Southern California Edison, aims to build a dynamic wildfire risk index using satellite imagery and machine learning. Motivated by increasingly severe fire seasons and the role of human and infrastructural triggers, the project employs Sentinel-2 satellite data and Multiple Endmember Spectral Mixture Analysis (MESMA) to assess vegetation health and fuel conditions. The project also integrates ML-Ops practices to automate and scale model training and deployment, making wildfire prediction more robust, reproducible, and production-ready.

Speaker Bio:
Daksha Singhal recently completed his Master’s in Computer Science at CU Boulder. At Earth Lab, he worked on a wildfire risk prediction project for Southern California Edison. Before grad school, he spent two years at Dish Networks as a backend engineer on services supporting Sling TV. He’s now working as a Java developer at Motion Industries.