Diabetic Macular Edema (DME) is one of the major causes of blindness in adults, primarily due to fluid accumulation in the macula derived from Diabetic Retinopathy (DR). In this paper, a Deep Learning (DL)-based diagnostic framework is developed to detect and identify various levels of DR severity, thereby facilitating early identification and management of DME. Our methodology begins with a comparison of several convolutional neural network (CNN) architectures for baseline modeling, including VGG 13, VGG 16, Resnet 18, Resnet 101, ResNeXt 50, EfficientNet B0, EfficientNet B3, EfficientNet B4, and ShuffleNet. Among these, the VGG-13 model was found to be the best performing model with an accuracy of 84.6% in identifying the severity of DR. Based on these outcomes, we introduce a Mixture of Experts (MoE) model that combines the capabilities of several task-specific models to increase the overall classification quality. This ensemble-based method improves the diagnostic accuracy to 93.7%, representing a 9.1% improvement over the baseline. These results underscore the potential of MoE in providing reliable DME diagnosis.