Global food security is increasingly threatened by the rising frequency and geographic spread of climate extremes such as droughts and heatwaves, which often trigger simultaneous crop failures across multiple agricultural regions, or "breadbaskets." These concurrent failures undermine food system resilience, causing significant production deficits, food shortages, price spikes, and political instability — threatening the sustainability of social-ecological systems. Despite such severe impacts, the spatiotemporal patterns of yield failure synchronizations and their atmospheric and climatic drivers remain poorly understood, representing a critical knowledge gap for global food security governance. The SAFE-Drivers project addresses this gap through an innovative multi-method approach integrating advanced statistical tools, machine learning, and physical atmospheric modeling, with a focus on maize, wheat, soybean, and rice. First, network theory will be applied to identify synchronizations of crop failures across ~20,000 political units over four decades of yield statistics. Second, causal discovery algorithms and explainable artificial intelligence, combined with Earth observations and climate reanalysis, will elucidate how atmospheric precursors and local climate factors drive these synchronized failures. Third, a Lagrangian atmospheric moisture and heat tracking framework (FLEXPART) will be used to trace the spatial propagation of climate extremes and their teleconnected agricultural impacts across global breadbaskets. The fellowship also provides interdisciplinary training in sustainability science, building the applicant's capacity for independent research leadership at the climate–agriculture interface. Overall, the project will substantially advance scientific understanding of synchronized crop failures, strengthen forecasting capabilities, and provide actionable insights for coordinated agricultural risk management — particularly benefiting food-insecure regions through improved early warning systems.