Cardiovascular Diseases (CVDs) continue to pose a significant global health concern with an alarming rise among younger populations. This study introduces a predictive framework using Tabular Network (TabNet), an advanced deep learning-based architecture tailored for structured data to detect CVDs. Unlike traditional neural networks, TabNet is specifically designed to process tabular data efficiently and its built-in attention mechanism also highlights the influential features, making it a white box deep learning architecture. The Behavioral Risk Factor Surveillance System (BRFSS) dataset was used, and class imbalance was addressed using Synthetic Minority Oversampling Technique (SMOTE). TabNet model achieved 90.26% accuracy, 89.98% F1-score, 92.58% precision, 87.52% recall and 96.33% Area Under the Curve-Receiver Operating Characteristic (AUC-ROC), outperforming traditional methods. TabNet’s built-in attention mechanism provides feature importance analysis, enhancing model interpretability. This study highlights TabNet’s potential in AI-driven healthcare, offering accurate, interpretable, and scalable solutions for proactive CVDs detection. Moreover, this framework shows potential for adoption in practical healthcare environments, supporting early diagnosis and prevention of CVDs through data-driven decision-making.