The explosive growth of online job portals has made manual resume screening an unsustainable bottleneck in modern recruitment. This paper presents a comprehensive AI-based Resume Analyser (AIRA) that automates candidate screening through a multi-stage pipeline integrating Natural Language Processing (NLP), Named Entity Recognition (NER), transformer based semantic embeddings (BERT/SBERT), and a weighted multi-criteria ranking engine. AIRA parses heterogeneous resume formats (PDF, DOCX, TXT), extracts structured entities— skills, education, experience, certifications, and languages—and matches candidates against job descriptions using both lexical and deep semantic similarity. A fine-tuned DistilBERT NER model achieves a macro-average F1 of 93.0% across eight entity types on a corpus of 5,000 annotated resumes spanning ten job domains. End-to-end candidate scoring surpasses keyword-based baselines by 22.9 F1 points and reduces recruiter screening time by 78%. An ablation study confirms the independent contribution of each pipeline module. Bias-mitigation mechanisms including PII anonymization and EEOC disparate impact testing are embedded in the design. Our results demonstrate that AIRA is a production-viable, fair, and highly accurate tool for the modern recruitment ecosystem.