This paper presents an Automated Resume Screening System (ARSS) that leverages Natural Language Processing (NLP) and Machine Learning (ML) techniques to intelligently parse, rank, and shortlist candidate resumes against a given job description. Traditional manual resume screening is time-consuming, inconsistent, and prone to human bias. The proposed hybrid approach combines TF-IDF vectorization, Word2Vec embeddings, and a fine-tuned BERT-based transformer model to extract semantic features from both resumes and job descriptions. Cosine similarity and a Support Vector Machine (SVM) classifier are employed for relevance scoring and candidate ranking. Experimental evaluation on a benchmark dataset of 5,000 resumes demonstrates a classification accuracy of 93.2%, precision of 93.6%, recall of 92.1%, and F1-score of 92.8%, outperforming existing baseline approaches. The system significantly reduces human screening effort by up to 78% while maintaining fairness and transparency in the shortlisting process.

Index Terms—Natural Language Processing, Resume Screening, TF-IDF, BERT, Word2Vec, Machine Learning, Cosine Similarity, SVM, Candidate Ranking.