Personalized learning has emerged as a crucial paradigm in contemporary education, yet current platforms frequently fail to provide suggestions for information relevant to each student’s requirements. In order to overcome this constraint, this work presents EduRAG, a domain-specific, AI-driven learning platform based on Retrieval-Augmented Generation (RAG). Students may submit their own study materials, such as lecture notes and textbooks, to EduRAG, which uses the provided materials alone to provide contextually appropriate answers and insights. The platform combines scalable serverless architecture with a sophisticated NLP algorithm to provide correct and timely replies. Three essential elements form the foundation of EduRAG’s system architecture: (1) processing documents utilizing Optical Character Recognition (OCR) and FAISS for embedding-based indexing; (2) a RAG pipeline that combines optimized language generation models with a high-performance retriever; and (3) Flexible study plans that let users customize their learning and give priority to particular subjects. With a concentration on STEM fields (mathematics, physics, and chemistry), the system’s initial implementation provides domain-specific accuracy and insights. An inventive blueprint-based query ranking system, thorough assessments of answer quality, and performance benchmarking against cutting-edge learning platforms are some of this work’s main achievements. The findings show that while managing several user requests simultaneously, EduRAG delivers improved engagement metrics, lower latency, and greater scalability. By lowering AI bias and implementing strong data protection safeguards, the platform also prioritizes ethical issues. This article highlights EduRAG’s potential as a game-changing tool in customized education by discussing its roadmap, which includes future integrations of multilingual support, gamification, and increased subject coverage.