Development of an Autonomous AI Interview Engine Using Sentence-BERT and Large Language Models for End-to-End Candidate Evaluation

The recruitment process in modern organisations is increasingly challenged by the growing volume of applications, manual resume screening inefficiencies, unconscious human bias, and the repetitive nature of initial technical interviews. Traditional Applicant Tracking Systems (ATS) primarily rely on keyword-based matching techniques, which often fail to capture the semantic relevance of a candidate’s skills, experience, and project work with respect to a given job description. As a result, qualified candidates may be overlooked due to variations in terminology or phrasing. This paper proposes the design and implementation of an Autonomous AI Interview Engine that performs end-to-end candidate evaluation by integrating semantic text understanding and conversational artificial intelligence. In the proposed system, Sentence-BERT (S-BERT) is utilised to generate dense semantic embeddings for both candidate resumes and job descriptions, enabling accurate similarity computation based on contextual meaning rather than lexical overlap. This semantic score forms the foundation of an intelligent screening mechanism that ranks candidates according to relevance and experience alignment. Beyond resume analysis, the system incorporates Large Language Models to conduct fully autonomous, real-time technical interviews through a voice-based interaction pipeline. The interview engine employs speech-to-text conversion to transcribe candidate responses, adaptive prompt engineering to generate context- aware follow-up questions, and text-to-speech synthesis to deliver human-like interviewer responses. A dedicated time-management module ensures consistent interview durations, while conversational state tracking allows the system to dynamically adjust question difficulty based on candidate performance. Upon interview completion, the system automatically evaluates the candidate using a structured scoring framework, assessing technical proficiency, communication skills, and project understanding. Experimental evaluation demonstrates that the proposed system can accurately rank candidates using semantic similarity, conduct coherent multi-turn technical interviews with low latency, and generate reliable evaluation reports. The results indicate that the Autonomous AI Interview Engine offers a scalable, objective, and efficient alternative to conventional recruitment screening and interviewing processes.