1,2Department of Pharmacognosy & Phytopharmacy, JSS College of Pharmacy, JSS Academy of Higher Education and Research, Ooty - 643 001, Nilgiris, Tamil Nadu, India.
3,4Department of Pharmaceutics, JSS College of Pharmacy, JSS Academy of Higher Education and Research, Ooty - 643 001, Nilgiris, Tamil Nadu, India.
The growing sophistication of Good Manufacturing Practice (GMP) requirements has highlighted the constraints inherent in conventional audit practices and especially their subjectivity, inconsistent interpretation, and fragmented monitoring of corrective measures. This paper explains the creation and multi-location testing of WinAI 2.0, a two-level artificial intelligence framework that was developed to standardise the interpretation of audits based on deterministic regulatory reasoning and domain-specific natural language processing. The system uses a weighted rule-engine based on regulatory clauses and contextual language modelling to transform unstructured observations of the audit into structured severity levels. A bias detection module is built into it and is used to compare the human and algorithmic scores to detect abnormal deviations, producing explainable severity-conflict indicators without disaggregating auditor opinion. The platform is also integrated with closed loop Corrective and Preventive Action (CAPA) automation, role-based access control, and secure cloud deployment to provide integrity and traceability of data. The V-model lifecycle (IQ/OQ/PQ/UAT) was used to validate five pharma manufacturing sites based on the GAMP5, 21C Part 11 and EU Annex 11 guidelines. One thousand three hundred and twenty audit observations have been analysed in order to test interpretive concordance and system reliability. The findings indicated high categorical consistency of severity ratings generated by AI and auditor (Cohen's kappa 0.88) and high numerical consistency of severity ratings (Pearson r = 0.94). The bias detection performance of the system was tested with the help of controlled bias-injection testing (n = 15), during which 13 cases out of the injected cases were detected. The role segregation was done in full and no data-integrity violations were noticed. These results indicate that such a hybrid, explainable AI-based method can be a considerable step to enhance consistency, reliability, and compliance in GMP auditing under validated circumstances.
Keywords: WinAI v2.0, Pharmaceutical auditing, Artificial Intelligence, CAPA management, RegTech, Validation, Quality 4.0.
How to cite this article: J JN, Dhanabal SP, D N, Karri VVSR. Advanced Validation and Multi-Site Implementation of WinAI: An AI-Driven Auditing and CAPA Management System for Pharmaceutical OSD Manufacturing Facilities. Int J Drug Deliv Technol. 2026;16(3): 281-289. DOI: 10.25258/ijddt.16.3.34
Source of support: Nil.
Conflict of interest: None