Cardiovascular disease remains a critical global health concern, necessitating early and accurate diagnosis to improve patient outcomes. Traditional auscultation methods rely on medical expertise, making them susceptible to diagnostic variability. This study introduces an AI-driven intelligent system for heart sound classification utilizing Wav2Vec 2.0 and Vision Transformer (ViT), two state-of-the-art deep learning architectures. Wav2Vec 2.0 processes raw waveform data, leveraging self-supervised learning to extract high-level audio representations, while ViT converts heart sound recordings into spectrograms for visual feature analysis. The research follows a systematic Knowledge Discovery in Databases (KDD) process, including data collection, augmentation, and feature extraction. Data augmentation techniques such as time stretching and pitch shifting enhance model generalization, addressing dataset limitations. Evaluation metrics, including accuracy, precision, recall, and F1 score, indicate superior performance, with both models achieving 99% classification accuracy, surpassing traditional approaches like CNN and SVM. The findings highlight the potential of transformer-based models in real-time automated cardiovascular diagnostics, paving the way for AI-assisted stethoscope applications in telemedicine and wearable health monitoring. Future research should focus on expanding datasets and optimizing computational efficiency to enhance model robustness across diverse populations.