Agricultural forecasting system
LLM-based advisory system backed by ML risk-forecasting models trained on Wisconet weather station data.
An LLM-based advisory system that helps growers reason about crop disease risk. The natural-language front end is backed by ML models that forecast risk from real-time Wisconet weather station data, so growers can ask questions in plain language and get model-grounded recommendations for spraying decisions.
- Live app: connect.doit.wisc.edu/ag_forecasting
- Source code: UW-Madison-DSI/ag_forecasting_app_v3
“There is a barrier of complexity of models that we can deliver to the masses. This code and tool help remove this barrier. They allow us to use very complex models, including advanced machine learning to improve accuracy, while keeping the usability and availability possible for the agricultural masses.”
— Damon Smith