The "WHEAT.AI" project aims to revolutionize agronomic intelligence by integrating machine learning and Earth observation to improve winter wheat cultivation in Switzerland. At its core is the development of WHEAT.AI, a transformer-based deep learning model that predicts Leaf Area Index (LAI) time series using weather, soil, and cultivation data. The project evaluates how these factors influence crop growth and yield, enhancing understanding of agricultural risk and resilience. WHEAT.AI will be validated against mechanistic models and field experiments to ensure robustness. Leveraging the multi-year MYWG dataset—including LAI observations, environmental variables, and detailed farming practices over eight years—WHEAT.AI bridges ecophysiology and remote sensing by embedding plant growth processes into Earth observation. The project’s work packages focus on model development, performance evaluation, and analysis of weather, soil, and management impacts on yield. Ultimately, WHEAT.AI introduces new applications for Earth observation, including improved data gap-filling and denoising, with broad implications for sustainable agriculture and large-scale crop monitoring.