TALENT TREK: ENHANCING INTERVIEW DECISIONS WITH CONVERSATIONAL AI

In the evolving landscape of recruitment, traditional hiring methods are increasingly inadequate for identifying top talent, particularly given the demands of modern, data-driven industries. This research introduces TALENT TREK, a modular, AI-powered Automated Human Resources (HR) Interview System designed to deliver fair, scalable, and multimodal candidate evaluations.

The system integrates real-time job data scraping, skill forecasting, semantic Natural Language Processing (NLP) based response analysis, and facial emotion recognition within a microservicesbased architecture optimized for concurrent processing. Leveraging a multi-engine speech recognition ensemble and the all-mpnet-base-v2 transformer for semantic evaluation, it achieves high correlation with human assessments while minimizing transcription and comprehension errors.

A custom Convolutional Neural Network (CNN) trained on FER2013 with domain-specific augmentations supports emotion classification, from which a novel Positive Confidence Score is derived. Multimodal data fusion enables adaptive weighting based on input quality, ensuring accurate composite scoring. Extensive testing demonstrates the system's potential to enhance transparency, consistency, and efficiency in enterprise-level hiring processes.