Benchmarking LLMs and AI-Driven Speech Processing for Interviews: An End-to-End Pipeline
Gopalsingh Saraf
Department of Information Technology
Pune Institute of Computer Technology
Pune, India gopalsaraf02@gmail.com
Rishikesh Revandikar
Department of Information Technology
Pune Institute of Computer Technology
Pune, India rishikeshrevandikar3110@gmail.com
Prathamesh Shriramwar
Department of Information Technology
Pune Institute of Computer Technology
Pune, India prathameshshriramwar100@gmail.com
Shivanjali Thorat
Department of Information Technology
Pune Institute of Computer Technology
Pune, India shivanjalithorat28@gmail.com
Mrs. S. A. Jakhete
Department of Information Technology
Pune Institute of Computer Technology
Pune, India sumeetra.kasat@gmail.com
Abstract—The competitive nature of today’s job market has amplified the challenges faced by both freshers and professionals, particularly those with limited practical experience, leading to increased anxiety, lack of confidence, and inadequate perfor- mance. This paper addresses the urgent need for personalized and effective interview preparation tools by introducing an AI- powered mock interview platform designed to simulate real- world scenarios and offer comprehensive, actionable feedback. Leveraging advanced AI models, including Llama 3.1 for dy- namic, context-aware simulations, Coqui for high-quality text-to- speech output, and Whisper for accurate speech-to-text process- ing, the platform delivers a full-featured solution for technical, behavioral, and HR interviews. Key features include real-time transcription, intelligent feedback, and in-depth performance assessments to improve communication skills, boost technical readiness, and build overall confidence. Our findings demonstrate that AI-driven solutions can significantly enhance the interview preparation process, offering scalable, unbiased, and highly effective experiences that empower candidates to succeed in a rapidly evolving job market.
Index Terms—Mock Interviews,Large Language Models, Intel- ligent Feedback, Performance Assessment, Llama 3.1, Whisper, Coqui
