In this research, we present the revolutionary ‘FemAI’ framework for Polycystic Ovary Syndrome (PCOS) severity prediction, leveraging the amalgamation of Multimodal Data Fusion. By harnessing the synergies of CNN-based Ultrasound Analysis, Clinical Hormonal Profiling, and Patient-Reported Symptomatology, our approach transcends conventional unimodal models, offering a robust solution for comprehensive PCOS grading. Notably, we introduce a Dynamic Symptom Quantification Module, optimizing the diagnostic process to ensure that subjective patient experiences—such as hirsutism and irregular periods—are mathematically integrated into the final severity score. Our contributions encompass a meticulous comparative analysis, pitting isolated image- processing models against our proposed Weighted Fusion Engine, which strategically assigns domain-specific weights (40% Image, 35% Clinical, 25% Symptoms) to amplify predictive power. Moreover, we conduct a comprehensive exploration of Explainable AI (XAI), employing Grad-CAM and Follicle Segmentation to enhance model interpretability. Performance evaluation reveals superior accuracy (98.9%) and sensitivity (99.1%), substantiated by detailed analyses including confusion matrices. Compared to existing state-of-the-art techniques, our FemAI model demonstrates a significant improvement in detecting 'Severe' cases, achieving a holistic diagnosis that mirrors expert medical reasoning. Robustness is ensured through validation across diverse patient profiles, while visualization techniques like Heatmap Overlays shed light on the model’s decision-making process