An Intelligent Multi-Modal Resume Screening System Using LLaMA-Based Semantic Embeddings, Weighted Skill Extraction, and Explainable Scoring

The growing volume of job applications poses a significant challenge for HR teams, making manual resume screening both time-consuming and error-prone. To address this, we propose the development of an intelligent, automated resume screening system based on LLaMA language models, designed to outperform traditional BERT-based approaches. The system leverages LLaMA’s advanced semantic understanding to generate embeddings of both resumes and job descriptions, using cosine similarity and a threshold-based retrieval mechanism to rank and filter candidates efficiently. To ensure transparency and build trust in automated decisions, the model incorporates Explainable AI (XAI) techniques, allowing HR professionals to understand the rationale behind each recommendation. This approach aims to significantly reduce screening time, increase accuracy, and support fairer, more interpretable hiring practices. Ultimately, the proposed system seeks to streamline recruitment workflows while upholding accountability and equity in candidate evaluation.