PREVUE.AI: A Web-Based Intelligent Mock Interview System Using Speech and Non-Verbal Analysis

Abstract

Interview preparation is a critical yet challenging phase for job seekers, as conventional preparation methods often lack realism, consistency, and objective evaluation. Traditional mock interviews depend heavily on human availability and subjective
feedback, making them difficult to scale and standardize. To address these challenges, this paper presents PREVUE.AI, an intelligent mock interview platform designed to simulate technical and HR interview scenarios using artificial intelligence. The
proposed system generates adaptive interview questions based on user responses, enabling a dynamic and context-aware interview flow. Spoken responses are captured and converted into text using automated speech recognition, allowing structured evaluation of answer relevance, clarity, and completeness. In addition to verbal analysis, the system incorporates basic non-verbal observation using MediaPipe-based facial landmark tracking to monitor visual indicators such as face presence and eye contact during interviews. This approach enhances interview realism without relying on complex emotional or sentiment detection. Interview results, scores, and session metadata are stored and visualized through an interactive dashboard, enabling users to track performance trends and improvement over multiple sessions. The proposed platform aims to provide an accessible, scalable, and objective interview preparation solution that closely reflects realworld interview conditions. By integrating adaptive questioning, speech-based evaluation, MediaPipe-driven visual observation, and performance analytics within a unified framework, PREVUE.  AI supports continuous learning and effective interview readiness.