AI Based Interview Preparation Platform Using MERN Stack and Generative AI

Background: At most Tier-2 and Tier-3 engineering colleges in India, students preparing for technical interviews are completely on their own. Study materials are not the problem — the internet has plenty. What is missing is any tool that actually simulates being asked a question and then tells you how well you answered it. Several tools exist for parts of this problem, but none of them works as a complete, integrated system a student can sit down and use. Objective: We describe the design, development, and preliminary evaluation of a web-based platform where students can complete full simulated interview sessions. The system generates questions, evaluates answers through a large language model, and stores performance data for tracking over time. No human is involved in any part of the assessment. Methods: The stack is MERN — MongoDB, Express.js, React.js, Node.js — with Google's Gemini API doing the question generation and answer evaluation. We also added voice input using the browser's built-in Web Speech API, email OTP verification via Resend, JWT-based authentication, and an analytics dashboard connected to MongoDB Atlas. The frontend runs on Vercel and the backend on Render. Results: We tested across six topics at three difficulty levels. Question generation averaged 2.1 seconds; answer evaluation came back within 3 seconds in more than 95% of attempts. A user study with 17 final-year B.Tech CSE students found that students who did three or more sessions on the same topic scored 28–32% higher by their last session compared to their first. 82% found the feedback helpful for spotting gaps. Conclusions: The platform fills a gap that we kept finding in the literature: a system that combines question generation, real-time AI evaluation, adjustable difficulty, and performance history in something free that runs in a browser. The improvement we saw among repeat users is encouraging, though the sample is too small to generalise from. A larger study across different institutions would be worth doing. Plain Language Summary: Most engineering students in India prepare for job interviews by reading notes and watching YouTube. That covers the knowledge side reasonably well. But interviews are not tests of knowledge — they are tests of whether you can explain your knowledge clearly while someone watches you. That is a different skill, and most students never get to practise it before campus placement season starts. We built a free website where you pick a topic like JavaScript or Data Structures, answer ten questions one by one, and get a score and written feedback from an AI for each answer. A small trial with 17 students from our college showed that scores on the same topic improved noticeably after two or three sessions, and most students said the feedback pointed them toward things they actually needed to review.