Voltage profile optimization is a critical challenge in maintaining the stability and efficiency of power systems. Traditional methods for controlling voltage levels, such as PID-driven automatic voltage regulators (AVRs), often encounter limitations due to dynamic load variations, system nonlinearities, and susceptibility to noise. This paper presents an advanced machine learning (ML)-based approach for enhancing voltage profile regulation in power systems. By integrating ML algorithms with AVR operations, the proposed model predicts and adapts to voltage fluctuations with greater precision, enabling enhanced stability, reduced system losses, and improved adaptability to varying load conditions. The methodology leverages a robust dataset for model training and applies proper data normalization to ensure consistent performance under real-time operating conditions. Simulation results, validated through quantitative metrics such as Mean Squared Error (MSE) and response time, reveal substantial improvements over conventional PID controllers, particularly in handling noisy inputs and dynamic scenarios. The proposed solution provides a scalable, intelligent, and practical framework for modern power grids, addressing key challenges such as system reliability, efficiency, and resilience.