In recent years, programming education has become an essential part of modern learning. However, evaluating student code manually is time-consuming and often inconsistent. To address this issue, this paper presents an AI-driven code analysis and feedback system that works using offline language models. The proposed system is capable of analyzing source code, identifying errors, and generating meaningful feedback similar to a human instructor. Unlike existing solutions that rely on cloud-based AI, this system operates completely offline using locally deployed models through the Ollama framework. This ensures data privacy, reduced dependency on internet connectivity, and faster response time. The system is designed to assist both students and educators by providing structured and easy-to-understand feedback. Overall, the proposed approach improves the efficiency, accuracy, and accessibility of programming education systems.This paper proposes a fully offline AI-driven code analysis and feedback system that integrates static program analysis with a locally deployed transformer-based language model to generate accurate, contextual, and pedagogically meaningful feedback The proposed system offers a secure, scalable, and cost-effective solution for programming education in resource-constrained academic environments.

This work demonstrates that it is possible to build an effective code analysis and feedback system without relying on cloud-based AI services. By combining static analysis with a locally deployed language model, the system achieves a balance between accuracy, usability, and privacy. The results suggest that such systems can play a meaningful role in programming education, particularly in environments where resources are limited. While there is room for further improvement, especially in handling complex code scenarios, the current implementation provides a solid foundation for future development.

[1][1] Z. Feng, D. Guo, D. Tang, N. Duan, X. Feng, M. Gong, L. Shou, B. Qin, T. Liu, D. Jiang, and M. Zhou, “CodeBERT: A Pre-Trained Model for Programming and Natural Languages,” Findings of the Association for Computational Linguistics: EMNLP 2020, pp. 1536–1547, 2020. doi: 10.18653/v1/2020.findings-emnlp.139 [2][2] C. Liu, S. Lu, W. Chen, D. Jiang, A. Svyatkovskiy, S. Fu, N. Sundaresan, and N. Duan, “Code Execution with Pre-trained Language Models,” Findings of ACL 2023, pp. 4984–4999, 2023. doi: 10.18653/v1/2023.findings-acl.308 [3][3] S. Xu and X. Yin, “Recommendation System for Privacy-Preserving Education Technologies,” Computational Intelligence and Neuroscience, vol. 2022, Art. no. 3502992, 2022. doi: 10.1155/2022/3502992 [4][4] J. I. del Valle and F. Lara, “AI-powered Recommender Systems and the Preservation of Personal Autonomy,” AI & Society, vol. 39, pp. 2479–2491, 2024. doi: 10.1007/s00146-02301720-2 [5][5] G. Diao, F. Liu, Z. Zuo, et al., “Privacy-aware and Efficient Student Clustering for Sport Training with Hash in Cloud Environment,” Journal of Cloud Computing, vol. 11, Art. no. 52, 2022. doi: 10.1186/s13677-022-00325-2 [6][6] I. A. Ismail and J. Aloshi, “Data Privacy in AI-Driven Education: An In-Depth Exploration into the Data Privacy Concerns and Potential Solutions,” in AI Applications and Strategies in Teacher Education, IGI Global, Jan. 2025, ch. 8, pp. 223–252. doi: 10.4018/979-8-3693-54438.ch008 [7][7] A. A. Salah, L. Colonna, and F. Florez-Revuelta (Eds.), Privacy-Aware Monitoring for Assisted Living: Ethical, Legal, and Technological Aspects of Audio- and Video-Based AAL Solutions, Cham: Springer, 2025. doi: 10.1007/978-3-031-84158-3 [8][8] R. Palle and K. C. R. Kathala, Privacy in the Age of Innovation: AI Solutions for Information Security, Berkeley, CA: Apress, 2024. doi: 10.1007/979-8-8688-0461-8 [9][9] B. Balami and J. Shakya, “Comparative Analysis of Transformer and CodeBERT for Program Translation,” Nepal Computer Science Research Journal, vol. 3, no. 1, 2024. doi: 10.3126/nccsrj.v3i1.72334 [10] D. Guo, S. Ren, S. Lu, Z. Feng, D. Tang, S. Liu, L. Zhou, N. Duan, A. Svyatkovskiy, S. Fu, M. Tufano, and S. K. Deng, “GraphCodeBERT: Pre-training Code Representations with Data Flow,” arXiv preprint, Sept. 2020. doi: 10.48550/arXiv.2009.08366 [11] X. Chen et al., “CodeTransFix: A Neural Machine Translation Approach for Context-Aware Java Program Repair with CodeBERT,” Applied Sciences, vol. 15, no. 7, Art. 3632, 2025. doi: 10.3390/app15073632 [12] S. Gupta and R. Mehta, “Automated Assessment Tools for Programming Education,” IEEE Access, vol. 10, pp. 12210–12220, 2022. doi: 10.1109/ACCESS.2022.3145120 [13] M. Tanaka and Y. Saito, “Ethical Challenges in AI-Enabled Learning Environments,” IEEE Transactions on Learning Technologies, vol. 16, no. 2, pp. 145–158, 2023. doi: 10.1109/TLT.2023.3245781 [14] F. Ahmed and S. Kumar, “Hybrid AI Approaches in Programming Feedback Systems,” IEEE International Conference on Artificial Intelligence and Data Science (AIDAS), 2023. doi: 10.1109/AIDAS.2023.10201456 [15] J. Li and Y. Zhao, “Offline AI Systems for Privacy-Aware Learning,” IEEE Access, vol. 9, pp. 80423–80433, 2021. doi: 10.1109/ACCESS.2021.3087651 [16] R. Johnson and T. Chen, “Adaptive Code Feedback Using Deep NLP Models,” Computer Education Research Journal, vol. 12, no. 4, pp. 221–234, 2022. doi: 10.1016/j.cerj.2022.05.003 [17] H. Park and J. Kim, “Leveraging LLMs for Automated Grading of Student Code,” IEEE Transactions on Education, vol. 67, no. 1, pp. 34–43, 2024. doi: 10.1109/TE.2024.3368741 [18] K. Patel, “Federated Learning for Educational AI Systems,” IEEE Access, vol. 12, pp. 30510–30520, 2024. doi: 10.1109/ACCESS.2024.3457623 [19] L. Zhang, “Offline AI Deployment for Data-Secure Institutions,” IEEE Cloud Computing, vol. 10, no. 3, pp. 56–63, 2023. doi: 10.1109/MCC.2023.3289011 [20] A. Singh and R. Das, “Automated Feedback in Programming Education: A Review,” IEEE Transactions on Learning Technologies, vol. 15, no. 5, pp. 922–931, 2022. doi: 10.1109/TLT.2022.3190005

© 2025 International Journal of Engineering and Techniques (IJET).

Digital Object Identifier (DOI)

IJET assigns a unique DOI to every accepted article via Zenodo for persistent linking and citation. Your article’s DOI: https://doi.org/{{doi}}

Open DOI About Zenodo DOI

Close