The study introduces a novel approach for classifying and predicting the progression of Age-related Macular Degeneration (AMD) using Optical Coherence Tomography (OCT) images and Multiple Instance Learning (MIL). AMD is a leading cause of vision impairment worldwide, making effective monitoring and treatment essential, particularly with anti-VEGF therapy. However, the increasing number of patients and the frequency of follow-up visits pose challenges for healthcare systems. This approach addresses two key tasks: (1) classifying changes between consecutive 2D OCT B-scans and (2) predicting disease progression within a 3-month period. For task 1, the model incorporates contextual information from adjacent B-scans and applies bidirectional cross-attention to learn time-dependent features. For task 2, a MIL-based architecture is used to identify the most significant slices within an OCT volume. The results demonstrated the effectiveness of the proposed methods. In task 1, the model achieved a mean score of 0.7488 across all evaluation metrics. For task 2, the mean score was 0.4478, reflecting the complexity of disease progression prediction. This approach offers improvements over baseline models and contributes to the development automated tools for AMD management, potentially easing the burden on ophthalmology services and improving personalized patient care.