Artificial Intelligence based communication tools such as Gmail Smart Reply and chatbots can generate quick responses, but they often fail to reflect a user’s personal writing style and emotional tone. The proposed AI Mirror system introduces a cognitive digital twin that learns user communication behavior and generates personalized, context-aware replies. The system integrates Gmail and Instagram APIs to collect conversation history. Messages are converted into embeddings using Sentence Transformers and stored in Chroma DB for contextual retrieval. When a new message arrives, relevant past conversations are retrieved and provided to a fine-tuned LLaMA-2 model for reply generation. A BERT-based classifier evaluates generated responses based on factual, emotional, and relational factors before sending or displaying them for user approval through a dashboard interface. The system is developed using Flask/Fast API and Tailwind CSS. Experimental evaluation showed improved personalization, response relevance, and communication efficiency while maintaining secure and adaptive learning.

This work presents the development of AI Mirror, an intelligent communication system that generates personalized and context-aware replies using Artificial Intelligence techniques. The system combines embeddings, vector databases, and Large Language Models to improve communication quality while understanding individual user interaction patterns. Experimental evaluation showed that the system can generate relevant and adaptive responses across platforms such as Gmail and Instagram. The proposed architecture also supports future improvements in personalization, automation, and smart communication services.

[1][1] Gupta, S., Mehta, A., & Rao, K. (2020). Emotion-aware conversational agents using hybrid neural networks. IEEE Transactions on Affective Computing, 11(3), 452–464. https://doi.org/10.1109/TAFFC.2020.2969098 [2] Li, M., & Xu, J. (2019). Contextual personalization in intelligent assistants. ACM Transactions on Interactive Intelligent Systems, 9(2), 1–18. https://doi.org/10.1145/3321448 [3] Singh, D., Patel, N., & Joshi, H. (2021). Cognitive digital twins for adaptive industrial intelligence. IEEE Access, 9, 58701–58712. https://doi.org/10.1109/ACCESS.2021.3070193 [4] Kim, R., & Lee, Y. (2022). Multimodal emotion synthesis for empathy-driven AI dialogue systems. Neurocomputing, 500, 73–85. https://doi.org/10.1016/j.neucom.2022.01.066 agents. [5] Sharma, P., & Deshmukh, K. (2020). Personality detection using machine learning for intelligent Procedia https://doi.org/10.1016/j.procs.2020.09.227 Computer Science, 176, 1954–1963. [6] Wang, L., Chen, X., & Zhou, M. (2023). Adaptive AI companions for healthcare and elderly support. IEEE Journal of Biomedical and Health Informatics, 27(2), 492–504. https://doi.org/10.1109/JBHI.2023.3264329 [7] Patel, N., & Mehta, A. (2021). Design frameworks for human-AI communication. AI and Society, 36(4), 1121–1132. https://doi.org/10.1007/s00146-020-01059-3 [8] Roberts, T., & Allen, J. (2024). Cognitive reinforcement learning for human-like decision making. Knowledge-Based Systems, 285, 111249. https://doi.org/10.1016/j.knosys.2024.111249 [9] Feng, Z., Guo, D., Tang, D., Duan, N., Feng, X., Gong, M., Shou, L., Qin, B., Liu, T., Jiang, D., & Zhou, M. (2020). CodeBERT: Findings of the Association for Computational Linguistics: EMNLP 2020, 1536 1547. https://doi.org/10.18653/v1/2020.findings-emnlp.139 [10] Liu, C., Lu, S., Chen, W., Jiang, D., Svyatkovskiy, A., Fu, S., Sundaresan, N., & Duan, N. (2023). Code execution with pre-trained language models. Findings of ACL 2023, 4984–4999. https://doi.org/10.18653/v1/2023.findings-acl.308 [11] Xu, S., & Yin, X. (2022). Recommendation system for privacy-preserving education technologies. Computational Intelligence and Neuroscience, Article ID 3502992. https://doi.org/10.1155/2022/3502992 [12] del Valle, J. I., & Lara, F. (2024). AI-powered recommender systems and the preservation of personal autonomy. AI & Society, 39, 2479–2491. https://doi.org/10.1007/s00146-023-01720 2 [13] Diao, G., Liu, F., Zuo, Z., et al. (2022). Privacy-aware and efficient student clustering for sport training with hash in cloud environment. Journal of Cloud Computing, 11, 52. https://doi.org/10.1186/s13677-022-00325-2 [14] Ismail, I. A., & Aloshi, J. (2025). 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 (pp. 223–252). IGI Global. https://doi.org/10.4018/979-8-3693 5443-8.ch008 [15] Salah, A. A., Colonna, L., & Florez-Revuelta, F. (Eds.). (2025). Privacy-aware monitoring for assisted living: Ethical, legal, and technological aspects of audio- and video-based AAL solutions. Cham: Springer. https://doi.org/10.1007/978-3-031-84158-3 [16] Palle, R., & Kathala, K. C. R. (2024). Privacy in the age of innovation: AI solutions for information security. Berkeley, CA: Apress. https://doi.org/10.1007/979-8-8688-0461-8 translation. [17] Balami, B., & Shakya, J. (2024). Comparative analysis of Transformer and CodeBERT for program Nepal Computer https://doi.org/10.3126/nccsrj.v3i1.72334 Science Research Journal, 3(1). [18] Guo, D., Ren, S., Lu, S., Feng, Z., Tang, D., Liu, S., Zhou, L., Duan, N., Svyatkovskiy, A., Fu, S., Tufano, M., & Deng, S. K. (2020). GraphCodeBERT: Pre-training code representations with data flow. arXiv preprint. https://doi.org/10.48550/arXiv.2009.08366 [19] Chen, X., et al. (2025). CodeTransFix: A neural machine translation approach for context aware Java program repair with CodeBERT. Applied Sciences, 15(7), 3632. https://doi.org/10.3390/app15073632 [20] Gupta, S., & Mehta, R. (2022). Automated assessment tools for programming education. IEEE Access, 10, 12210–12220. https://doi.org/10.1109/ACCESS.2022.3145120 [21] Tanaka, M., & Saito, Y. (2023). Ethical challenges in AI-enabled learning environments. IEEE on Learning https://doi.org/10.1109/TLT.2023.3245781 Technologies, 16(2), 145–158. [22] Ahmed, F., & Kumar, S. (2023). Hybrid AI approaches in programming feedback systems. In IEEE International Conference on Artificial Intelligence and Data Science (AIDAS). https://doi.org/10.1109/AIDAS.2023.10201456 [23] Li, J., & Zhao, Y. (2021). Offline AI systems for privacy-aware learning. IEEE Access, 9, 80423–80433. https://doi.org/10.1109/ACCESS.2021.3087651 Education [24] Johnson, R., & Chen, T. (2022). Adaptive code feedback using deep NLP models. Computer Research https://doi.org/10.1016/j.cerj.2022.05.003 Journal, 12(4), 221–234. [25] Park, H., & Kim, J. (2024). Leveraging LLMs for automated grading of student code. IEEE Transactions on Education, 67(1), 34–43. https://doi.org/10.1109/TE.2024.3368741 [26] Patel, K. (2024). Federated learning for educational AI systems. IEEE Access, 12, 30510 30520. https://doi.org/10.1109/ACCESS.2024.3457623 [27] Zhang, L. (2023). Offline AI deployment for data-secure institutions. IEEE Cloud Computing, 10(3), 56–63. https://doi.org/10.1109/MCC.2023.3289011 Transactions [28] Singh, A., & Das, R. (2022). Automated feedback in programming education: A review. IEEE on Learning https://doi.org/10.1109/TLT.2022.3190005 Technologies, 15(5), 922–931. [29] Zheng, Z., & Guo, D. (2024). From digital human modeling to human digital twin: Framework and perspectives in human factors. Chinese Journal of Mechanical Engineering, 37(3). https://doi.org/10.1186/s10033-024-00998-7

© 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