• Designed and built a retrieval￾augmented pipeline matching candidates to job descriptions end to end. • Architected a hybrid retrieval layer combining BM25 keyword search with dense semantic embeddings over Elasticsearch and MongoDB, improving ranking relevance over either method alone. • Built NLP pipelines with HuggingFace Transformers and Word2Vec to compute semantic similarity between job postings and candidate profiles. • Exposed the system through a FastAPI backend with OAuth2 / JWT authentication; containerized the full stack with Docker for reproducible deployment.