ReFlow: Agentic AI-Based Self-Healing Data Pipeline for Data Quality Monitoring and Automated Recovery

This paper presents ReFlow, a proposed agentic AI-based self-healing data pipeline for data-quality monitoring and controlled recovery. The framework combines rule-based validation, Gemini-assisted diagnosis, approved repair functions, and independent revalidation. It focuses on structured Amazon product data, including missing values, inconsistent numeric formats, invalid prices, ratings, and duplicate candidates. The design preserves original records and maintains an incident history to make corrections traceable. Uncertain recommendations and unresolved issues are referred for human review. A Streamlit interface supports data upload, inspection of repairs, and comparison of original and processed datasets. The main contribution is a coordinated workflow that connects error detection with diagnosis, correction, and verification. Comparative evaluation remains pending; therefore, improvements in repair correctness, recovery time, or manual effort cannot yet be confirmed. Future work will evaluate these outcomes against a rule-based baseline and explore additional data sources, database integration, and streaming support.

[6] N. Sambasivan et al., “‘Everyone Wants to Do the Model Work, Not the Data Work’: Data Cascades in High-Stakes AI,” in Proceedings of the CHI Conference on Human Factors in Computing Systems, 2021. DOI[7] Y. Tashiro, J. Song, Y. Song, and S. Ermon, “CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputation,” in Advances in Neural Information Processing Systems (NeurIPS), 2021. Paper[8] J. Xu, H. Wu, J. Wang, and M. Long, “Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy,” in International Conference on Learning Representations (ICLR), 2022. Paper[9] J. Wei et al., “Chain-of-Thought Prompting Elicits Reasoning in Large Language Models,” in Advances in Neural Information Processing Systems (NeurIPS), 2022. Paper[10] S. Yao et al., “ReAct: Synergizing Reasoning and Acting in Language Models,” in International Conference on Learning Representations (ICLR), 2023. Paper[11] N. Shinn et al., “Reflexion: Language Agents with Verbal Reinforcement Learning,” in Advances in Neural Information Processing Systems (NeurIPS), 2023. Paper[12] A. Madaan et al., “Self-Refine: Iterative Refinement with Self-Feedback,” in Advances in Neural Information Processing Systems (NeurIPS), 2023. Paper[13] T. Schick et al., “Toolformer: Language Models Can Teach Themselves to Use Tools,” in Advances in Neural Information Processing Systems (NeurIPS), 2023. Paper[14] X. Wang et al., “Self-Consistency Improves Chain of Thought Reasoning in Language Models,” in International Conference on Learning Representations (ICLR), 2023. Paper