An ai driven mock interview system with real-time proctoring and automated feedback generation

Abstract

The present invention is related to an AI driven mock interview system with real-time proctoring and automated feedback generation. The invention enables users to give mock interviews conducted by a voice agent for preparation of an actual job interview with personalized feedback at the end of the interview with strict proctoring measures to avoid malpractice. The entire system is embedded in a high-speed computing device. Google’s Speech-to-Text and Text-to-Speech APIs are used. Instead of YOLO based pretrained models and other conventional machine learning based models for proctoring, a novel adaptive calibration-based computer-vision solution is proposed. Instead of using traditional heavyweight models like BERT for NLP tasks, the lightweight all-MiniLM-L6-v2 model is used for driving speed and efficiency, giving similar results as compared to heavyweight NLP models. Due to an embedded high-speed computing device, since all processing occurs locally, the system can detect anomalies and prevent malpractices in real time without relying on internet connectivity, with low power requirements, making it ideal for continuous proctoring without excessive energy usage.