Modern employment requires resumes to be matched to job descriptions, and this can be a time-consuming process. We introduce Resume Refactorer, an artificial intelligence system that utilizes Retrieval-Augmented Generation (RAG), semantic similarity, and Large Language Models to optimize resumes from job descriptions (JD) automatically. Our system’s final iteration utilizes DeepSeek R1, an open-source high-performance model available through OpenRouter, for context-aware resume editing based on semantic retrieval from ChromaDB and SentenceTransformers. Users deal with a modular toggle-based interface and get Applicant Tracking System(ATS) scores calculated through ATS Utilities 3.3.2. Previous versions with local LLaMA 3.2 and fine-tuned TinyLlama guided system development but were held back by context handling. Measurement with cosine similarity and ATS scores on 10 job positions shows improvements in resume-JD similarity consistently. This research traces the progression and enhancement of Resume Refactorer into a retrieval-grounded, user-controlled, and scalable resume optimization pipeline.