This paper presents Beacon, a voice assistant designed to improve the accessibility of Windows PCs for visually impaired users. Beacon leverages the PhoBERT model, a state-of-the-art Vietnamese language processing tool, to accurately recognize and process voice commands. The system offers essential functions such as music playback, news reading, and volume control through voice interaction, minimizing the need for screen reliance. Built with FastAPI, Beacon ensures efficient processing and fast response times, achieving up to 90% accuracy in recognizing user intent. This research addresses the gap in voice assistants for non-English languages, providing an inclusive solution for visually impaired users, particularly in Vietnamese-speaking regions. The paper also outlines the system architecture, user interaction design, and results from comprehensive testing, demonstrating the potential of Beacon to empower visually impaired individuals with greater digital independence.