An AI-Powered Real-Time Voice-Based Interview Platform for Adaptive Interview Practice

Interview preparation is a critical factor in recruitment success; however, most existing preparation tools rely on static, text-based interactions that do not accurately represent real interview environments. While some platforms provide live mock interviews, their dependence on human interviewers makes them costly, time-consuming, and difficult to scale. This paper proposes an AI-driven real-time voice-based interview platform designed to simulate realistic interview scenarios through intelligent conversational interaction. The system dynamically generates and adapts interview questions based on candidate responses using large language models. Real-time voice communication is achieved through speech-totext and text-to-speech technologies, enabling natural interaction between the candidate and the AI interviewer. The platform is implemented using Next.js for frontend development, Firebase Authentication and Firestore for secure user management and data storage, Vapi for real-time voice interaction, and Google Gemini for intelligent question generation, response analysis, and feedback. By automating interview conduction and evaluation, the proposed solution reduces dependency on human interviewers while maintaining interview realism. The system offers a scalable, cost-effective, and interactive approach to interview preparation.