Accurate forecasting of pandemic dynamics is a critical component of effective public health planning and policy formulation. In this study, we introduce the Fairness-Aware Mobility-Integrated Transformer (FAMIT), a novel deep learning framework designed to capture the complex interplay between epidemiological, demographic, policy, mobility, and vaccination factors in modeling disease transmission. FAMIT incorporates a mobility-modulated attention mechanism that dynamically adapts to variations in population movement and non-pharmaceutical interventions, thereby improving temporal responsiveness. To address equity concerns, the framework integrates a fairness-aware loss function that mitigates prediction disparities across vulnerable demographic subgroups. Using a large-scale dataset of more than 429,435 records spanning 255 global locations, FAMIT consistently outperforms state-of-the-art benchmark models in both predictive accuracy and robustness, particularly during outbreak surges. Furthermore, its use of Monte Carlo Dropout provides tighter and more informative uncertainty estimates, enhancing interpretability. The findings underscore FAMIT’s potential to deliver reliable, equitable, and interpretable pandemic forecasts, thereby contributing to the development of resilient and sustainable public health infrastructures.