Agentic AI Base E-Learning System

Abstract

In recent years, digital learning has become very common, but most e-learning platforms still depend on pre-designed content and do not support user-provided study materials effectively. Students often have to read large textbooks without any structured guidance, which makes learning time-consuming and less interactive. To solve this problem, this project presents an Agentic AI-Based E-Learning System that converts textbook PDFs into structured and easy-to-understand learning content. The system allows users to upload any book in PDF format. After uploading, chapters are defined using page ranges, and the system extracts the content for each chapter. The extracted text is cleaned to remove unnecessary data such as unwanted symbols and formatting issues. Based on this cleaned content, the system identifies important topics and provides simple explanations to help users understand the concepts more easily In addition, the system generates multiple-choice questions for each topic so that users can test their understanding immediately after learning. The quiz results are stored in a database, allowing users to track their performance over time. A chatbot feature is also included to help users ask questions related to the topic and get quick answers The system is developed using Flask for backend processing, PyMuPDF for PDF handling, and AI-based models for topic extraction, explanation, and quiz generation. This approach reduces manual effort and makes learning more interactive, organized, and user-friendly.

Introduction

The Agentic AI-Based E-Learning System is designed to transform traditional PDF-based learning materials into an interactive and structured learning platform. Existing e-learning systems mainly depend on predefined courses and content, while students studying from large PDF textbooks often face difficulties in identifying important topics, preparing notes, and testing their understanding. The proposed system solves these problems by using Artificial Intelligence to automatically organize and enhance learning content.

The system allows users to upload any PDF textbook and converts it into a structured learning format. It extracts chapter-wise content, identifies important topics, generates simple explanations, creates multiple-choice quizzes, stores performance records, and provides an AI chatbot for answering topic-related questions.

The literature review highlights that existing AI education systems use technologies such as GPT models, BERT, Retrieval-Augmented Generation (RAG), and local AI models for content extraction, question generation, and chatbot-based learning. However, many existing solutions depend heavily on cloud services, require continuous internet access, or lack complete integration of textbook processing, learning assistance, and performance tracking. The proposed system addresses these gaps through a modular and partially offline-capable architecture.

The system workflow includes:

• PDF Upload Module: Users upload textbooks in PDF format.
• Chapter Processing Module: Content is divided into chapters using user-defined page ranges.
• Text Extraction and Cleaning: Extracts and prepares readable content from PDFs.
• AI Topic Extraction: Identifies important learning topics from chapters.
• Explanation Generation: Provides simplified AI-generated explanations.
• Quiz Generation: Creates multiple-choice questions for self-assessment.
• Result Tracking: Stores quiz scores and learning progress.
• Chatbot Module: Allows users to ask questions and receive instant responses.

The architecture consists of five major layers:

1. User Layer – Provides the web interface for uploading books and learning.
2. Application Layer – Handles system logic using the backend.
3. Processing Layer – Performs PDF extraction, cleaning, and chapter organization.
4. AI Layer – Generates topics, explanations, quizzes, and chatbot responses.
5. Data Layer – Stores processed content and user performance using SQLite.

The implementation uses technologies such as Flask, PyMuPDF, AI models, and SQLite database to create a complete learning pipeline. Testing with different PDF books showed successful conversion of raw documents into structured learning materials. The quiz system generated effective assessments, and stored results helped users monitor their progress.

Advantages:

• Converts large PDF books into organized learning content
• Reduces manual study effort
• Provides topic-based explanations
• Generates quizzes for knowledge testing
• Tracks learning performance
• Offers an interactive chatbot experience

Limitations:

• AI features may require internet connectivity
• Limited support for scanned or complex PDFs
• Accuracy depends on PDF quality
• Processing time increases for large documents

Future Scope:

• Improve offline AI processing capability
• Add better support for scanned documents using OCR
• Enhance AI models for more accurate explanations
• Improve handling of noisy or low-quality content
• Add advanced personalized learning recommendations

Conclusion

The proposed Agentic AI-Based E-Learning System successfully converts traditional PDF-based study material into a structured and interactive learning experience. The system simplifies the learning process by extracting chapters, identifying key topics, and providing easy-to-understand explanations. The integration of quiz generation helps users test their knowledge immediately, while the result tracking feature allows them to monitor their progress. Overall, the system reduces manual effort and makes learning more organized and efficient. Although the system performs well, it still depends on internet connectivity for AI features and has limited support for complex scanned PDFs. These limitations can be improved in future by integrating local AI models and better OCR techniques. In conclusion, the system demonstrates how AI can be effectively used to enhance digital learning and provide a more interactive and user-friendly educational platform.

References

[1] M. Kumar, S. Sarvajit Visagan, T. Mahajan, A. Natarajan, and P. S. Sreeja, \"Enhanced Sign Language Translation Between American Sign Language and Indian Sign Language Using LLMs,\" IEEE Access, vol. 13, pp. 156270-156XXX, 2025, doi: 10.1109/ACCESS.2025.3595943. [2] .A. Kumar and R. Sharma, “AI-Based Quiz Generation from Lecture Content Using GPT Models,” International Journal of Educational Technology, vol. 10, no. 2, pp. 45–52, 2022. [3] S. Patel, M. Verma, and K. Shah, “Offline Question Generation Using BERT for Educational Content,” Journal of Artificial Intelligence in Education, vol. 12, no. 1, pp. 78–85, 2023. [4] Y. Li, H. Chen, and X. Wang, “Retrieval-Augmented Generation for Document-Based Question Answering Using LLaMA and FAISS,” IEEE Access, vol. 11, pp. 102345–102356, 2023. [5] P. Gupta and S. Mehta, “Local Deployment of RAG Systems Using Ollama and ChromaDB for Educational Applications,” International Conference on Machine Learning Applications, pp. 210–215, 2024. [6] T. Gao, L. Sun, and J. Liu, “A Survey on Retrieval-Augmented Generation for Large Language Models,” ACM Computing Surveys, vol. 56, no. 3, pp. 1–36, 2023. [7] Z. Yu, Q. Zhang, and L. Zhao, “RAG-Based Chatbots in Education: A Comprehensive Survey (2023–2025),” Education and Information Technologies, vol. 29, pp. 1123–1145, 2025. [8] F. Lang and E. Gurpinar, “Context-Aware Educational Chatbots Using Retrieval-Augmented Generation,” IEEE Transactions on Learning Technologies, vol. 17, no. 1, pp. 55–63, 2024. [9] W. Zhao, Y. Lin, and K. Huang, “Enhancing Retrieval-Augmented Generation with Reinforcement Learning,” Proceedings of the AAAI Conference on Artificial Intelligence, pp. 9876–9883, 2024.

Copyright

Copyright © 2026 Rushikesh Bhalerao, Saurabh Ghangale , Suraj Jadhav , Suraj Kotkar , Om Mankar . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.