With the rapid rise of electric vehicle (EV) adoption in 2024-2025, ensuring reliability and serviceability is a major challenge. Traditional reactive or scheduled maintenance cannot adequately prevent sudden subsystem failures in batteries, motors, or power electronics, which cause downtime, safety risks, and higher costs. This study presents an Internet of Things (IoT)-based predictive maintenance framework for smart EVs, emphasizing real-time fault prediction and anomaly detection. The framework integrates sensor networks, cloud analytics, and machine learning models to anticipate failures before escalation. Key components include: (i) monitoring of the Battery Management System (BMS), traction motor, and thermal modules; (ii) a machine learning pipeline using Long Short-Term Memory (LSTM) for time-series forecasting and Random Forest for fault classification; and (iii) a scalable, cloud-enabled architecture for fleet-wide reliability. A prototype drivetrain instrumented with voltage, current, temperature, and vibration sensors transmitted data via a Raspberry Pi gateway over MQTT. The LSTM achieved above 92% accuracy in thermal forecasting, while the Random Forest reached up to 96% accuracy and balanced precision-recall above 92% with cross-validation. Results demonstrated a 30% reduction in unplanned service events and up to 8% efficiency gain, confirming IoT-driven predictive maintenance as a viable strategy for smart EVs