The Challenge
The client needed to proactively predict the pre-harvest risk of Fusarium mycotoxins in its maize supply chain across approximately 26,000 sourcing locations in France and Poland. This required an automated, weekly process to ingest, combine, and aggregate complex historical weather data from NASA and seasonal forecast data from ClimateAI, which was impossible with their existing systems.
Our Solution
We built a fully automated, scalable data pipeline using Azure Data Factory and Azure Databricks. The solution ingests data from both NASA and ClimateAI APIs, processing it against a geospatial grid of 26,000 H3 cells. The pipeline performs a "smart union" of historical and forecast data, runs it through a machine learning model, and automatically generates dual-format reports: a comprehensive regional heatmap for data science and a targeted supplier-specific report for business stakeholders.
Results & Impact
Delivered a fully automated weekly pipeline that provides actionable pre-harvest risk predictions. This enables business stakeholders to make informed, proactive sourcing decisions to ensure crop quality, while simultaneously providing the data science team with comprehensive regional data and forecast accuracy metrics for continuous model improvement.