ChaiMetrics
An ML platform for KTDA smallholder tea farmers spanning 240 farms and 15 factories. Yield forecasting runs on gradient-boosted trees and SARIMA time series, while a graph neural network models the farm-factory relational structure directly — treating collection routes and processing dependencies as a graph rather than flattening them into tabular features. An LLM narration layer translates model output into plain-language guidance for farmers with no ML background.
Repo ↗Read the full write-up
This is a placeholder for the full blog post content.
The actual content would be in Markdown format and converted to HTML.