InterviewAI: Real-time questions generator using LLM

Soon, Chun Hong (2025) InterviewAI: Real-time questions generator using LLM. Final Year Project, UTAR.

Preview

PDF Download (4Mb) | Preview

Abstract

InterviewAI is an advanced AI-driven platform designed to transform the recruitment process by integrating emotion detection, automated CV analysis, and dynamic question generation powered by large language models (LLMs). The primary goal of this Final Year Project (FYP2) is to enhance recruitment efficiency, fairness, and personalization by extracting critical information from CVs, analyzing candidates’ real-time emotional states, and generating tailored interview questions based on these insights. The system employs a sophisticated combination of machine learning models, including a Convolutional Neural Network (CNN) for real-time emotion detection from video feeds and the Llama3 model for context-aware question generation, seamlessly integrated into a unified framework. Leveraging Artificial Intelligence for data processing and Human-Computer Interaction principles for user-centric design, the methodology ensures robust handling of multimodal data, enabling the system to adapt dynamically to each candidate’s emotional and professional profile. Compared to its initial development in FYP1, InterviewAI in FYP2 has been significantly refined to improve accuracy in emotion detection, enhance CV extraction capabilities, and optimize question relevance through iterative model training and user feedback. Final results demonstrate the system’s ability to reduce recruitment bias, streamline the interview process, and provide HR professionals with an intelligent tool that adapts to individual candidate responses, thereby fostering a more inclusive and equitable interview experience. By alleviating the administrative burden on recruiters and promoting objective evaluations, InterviewAI showcases substantial potential to revolutionize modern recruitment practices, making them more efficient, unbiased, and tailored to the unique needs of each candidate.

Actions (login required)