A Data-Driven and Modular Flask-Based Architecture for Secure and Intelligent Programming Education Powered by LLMs
Keywords:
Flask Web Framework, Large Language Models (LLMs), LLaMA 2, Dynamic Question Generation, Secure Code ExecutionAbstract
In this paper, the research is about a modular, AI data-driven programming education platform developed using the Flask web framework and integrated with the LLaMA 2 large language model (LLM) to deliver dynamic, personalized learning experiences. The system combines real-time question generation, contextual feedback, and secure code execution through Docker containerization to ensure safe and isolated code evaluation across multiple programming languages, including Python, C, and C++. Architecture supports adaptive learning by analyzing user submissions and providing feedback on syntax, logic, efficiency, and coding style. Performance evaluation demonstrates that the system maintains optimal response times and throughput for up to 70 concurrent users, with CPU usage remaining below 80% and average response times under 300 ms. Beyond this threshold, resource utilization increases, and error rates rise, highlighting the need for future load balancing and optimization strategies. User testing further confirms high learner engagement and effectiveness, with over 85% of participants reporting improved understanding and satisfaction with real-time AI feedback. The platform’s modular design enables seamless integration of future enhancements, including support for additional languages, learning management system (LMS) interoperability, and gamification features. These results validate the proposed system as a secure, scalable, and intelligent solution for next-generation programming education.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2025 Journal of Data Science

This work is licensed under a Creative Commons Attribution 4.0 International License.