Paper Title
Prep-Wise: Smarter Prep with Every Paper

Abstract
This research work is relevant to the current scenario in the education sector, focusing on the challenges of organizing and searching academic resources and assisting in efficient exam preparation. The current university management framework provides only limited functionality in terms of administrative assistance, attendance tracking, and viewing grades. However, they do not provide sophisticated support in the form of intelligent document organization, semantic searching, and AI- assisted exam preparation tools. To overcome these limitations, this research work aims to develop a holistic framework that com- bines the capabilities of Retrieval-Augmented Generation (RAG), Optical Character Recognition (OCR), vector semantic search, and generative models. The Smart Academic Assistant Portal provides a two-fold Previous Year Question (PYQ) Manager and an AI-assisted Sample Paper Generator, using a modified Tesseract-NER pipeline for metadata extraction from PDF files, PostgreSQL for storing relational data, and Qdrant vector embeddings for context-dependent searching. The system uses few-shot prompting with local large language model inference (llama3.2:3b via Ollama) to generate difficulty-level normalized mock question papers based on Bloom’s Taxonomy. The system comes with a Next.js frontend for responsive user interaction and a FastAPI backend for scalability using asynchronous request processing. The performance analysis shows that the semantic search functionality decreases false positives by as much as 45% com-pared to keyword-based filtering alone, and the few-shot prompted question generation functionality provides 71% immediate usability with logically consistent and pattern-following questions. Keywords - Retrieval-Augmented Generation, Semantic Search, Optical Character Recognition, Question Generation, Academic Repositories, Vector Embeddings, Large Language Models