The rising global prevalence of diabetes necessitates noninvasive, real-time, patient-centric monitoring solutions. This study presents an integrated system that combines a custom-designed wearable device with a volatile organic compound (VOC) sensor for acetone-based diabetes detection, a machine learning (ML)-based predictive model, and a web-enabled platform for continuous engagement in healthcare. The wearable unit monitors vital physiological parameters, including heart rate, SpO₂, temperature, blood pressure, and activity, whereas the VOC sensor estimates the breath acetone concentration for noninvasive glucose level indication. A machine learning pipeline employing five classifiers–KNN, Logistic Regression, Decision Tree, Random Forest, and Gradient Boosting Classifier (GBC)—was evaluated, with GBC achieved the highest predictive performance (accuracy: 92.2%, ROCAUC: 0.993). The system includes a secure, responsive web interface developed using Streamlit and Firebase, offering real-time visualization, patient–doctor communication, and an AI-powered chatbot built using the Gemini API for intelligent health assistance.