Retrieval-Augmented Generative AI Study Assistant for Indian Academic Textbooks

Objective: To develop and deploy a generative AI-powered study assistant that provides accurate, context-aware Q&A over prescribed Indian textbook content using retrieval-augmented generation.

Many Indian students face challenges in accessing concise, relevant explanations and answers from voluminous academic textbooks, especially in tier-2 and rural colleges where live faculty support is limited. This gap affects exam preparedness, conceptual clarity, and equal access to quality education, impacting millions of learners preparing for board, university, and competitive examinations.

This project proposes an end-to-end engineering solution: a web/mobile study assistant application that leverages retrieval-augmented generation (RAG) using large language models (LLMs) and custom vector databases built from NPTEL, NCERT, and Mumbai University prescribed textbooks. The system combines natural language question understanding, semantic search over textbook embeddings, and generative answer synthesis, ensuring factual accuracy by grounding outputs in retrieved text passages.

Key features include: textbook ingestion pipeline with OCR and data cleaning, semantic search and retrieval using FAISS or Pinecone, integration with open-source LLMs (Llama 2, Falcon, or GPT-3.5 via API), user-facing Q&A interface, citation of textbook passages in answers, administrative dashboard for usage analytics, and deployment on cloud infrastructure. The final working model includes a live demonstration with sample university textbooks and real-time user queries.

This solution can be scaled to support multiple Indian languages and curricula, democratizing access to high-quality, context-aware study assistance, reducing dependence on coaching, and supporting inclusive digital education initiatives nationwide.