What question did this study set out to answer?

This research aims to develop and evaluate an AI-driven platform that personalizes competitive examination preparation for students.

May 21, 2026Open Access

PRISM - An Agentic RAG-Based Platform for Competitive Examination Preparation with Personalized Learning and Real-Time Collaboration

Key Points

This research aims to develop and evaluate an AI-driven platform that personalizes competitive examination preparation for students.
Developed an AI platform using agentic retrieval-augmented generation techniques.
Indexed 19,776 embeddings from over 40 academic PDFs for effective resource retrieval.
Implemented real-time collaborative features and adaptive study planning using a cross-device system.
Achieved sub-5ms retrieval latency on standard CPU hardware.
Demonstrated improved student engagement and learning outcomes in preparation for JEE and NEET exams.

Abstract

PRISM (Personalized Resource and Intelligent Study Monitor) is an AI-powered competitive examination preparation platform designed for JEE Mains, JEE Advanced, and NEET students. The system implements a five-node agentic Retrieval-Augmented Generation (RAG) pipeline using LangGraph, LangChain, FAISS, and the Ollama llama3.2 language model. A FAISS vector store indexes 19,776 embeddings from 40+ academic PDFs including NCERT textbooks, previous year question papers, and supplementary materials using the all-MiniLM-L6-v2 sentence transformer. The platform integrates PYQ-mode quiz generation, NCERT line-by-line reading with global content caching, real-time competitive battle rooms via Socket.IO, adaptive study planning, and a cross-device leaderboard backed by MongoDB Atlas. Deployed locally without cloud API dependency, the system achieves sub-5ms retrieval latency on commodity CPU hardware.

PRISM - An Agentic RAG-Based Platform for Competitive Examination Preparation with Personalized Learning and Real-Time Collaboration

Key Points

Abstract

Cite This Study