PrepPilot: An AI-Based Personalized Mock Interview Platform for Students and Job Seekers

Abstract

The transition from academic learning to professional employment is critically dependent on interview performance, yet most fresh graduates lack access to consistent, personalized, and realistic mock interview opportunities. Conventional preparation methods such as reading question lists, watching online videos, or solving coding problems on platforms like LeetCode and HackerRank build theoretical knowledge but fail to simulate the pressure, communication demands, and personalized questioning of a real interview. Existing AI-assisted tools such as Google Interview Warmup offer role-based practice but lack deep resume-level personalization, while peer-to-peer platforms like Pramp and Interview.io depend on human availability, limiting accessibility. Furthermore, many recruitment-focused systems such as Jobma and myInterview are designed for employers rather than for candidate self-improvement. This research gap motivates the development of PrepPilot. PrepPilot is a web-based platform developed using the MERN stack, which provides an end-to-end personalized interview preparation workflow. The system allows users to securely register and authenticate using JSON Web Tokens. Upon uploading a resume in PDF format, a PDF processing module extracts textual content which is then passed as context to the GroqCloud LLM API for generating highly relevant technical questions based on the candidate's own skills, projects and experience. The candidate then participates in a live mock interview using microphone and webcam integration enabled by the Web Speech API and browser media APIs. Spoken responses are transcribed and evaluated by the AI engine to generate question-wise scores and feedback. At the conclusion of the session, the system aggregates the evaluation to produce a comprehensive performance report comprising overall score, recommendation, strengths, weaknesses, actionable suggestions, and detailed feedback. Experimental testing with sample resumes for roles such as Full Stack Developer demonstrated successful end-to-end execution of resume extraction, contextual question generation, voice-based interaction, and automated report generation for the tested sample cases. These observations demonstrate technical feasibility of the implemented workflow; they do not establish psychometric validity, hiring accuracy, or generalizability to diverse populations. Prior research indicates that AI-supported interview training can provide scalable practice, while automated interview scoring requires careful validation of reliability, validity, and potential bias.