A FHIR-Native Microservices Architecture for an Integrated AI-Powered Health Assistant (AI-PHA): Design, Validation, and Regulatory Compliance in the Indian Context

Modern Electronic Health Record (EHR) systems frequently exhibit fragmentation, hindering interdisciplinary data flow and increasing clinical friction. This paper proposes the design and evaluation framework for an AI-Powered Health Assistant (AI-PHA) system, built on a secure, HL7 FHIR-native microservices architecture. The AI-PHA utilizes distinct, role-specific portals for Physicians, Patients, and Administrators, mitigating usability challenges inherent in generalized systems. A core component is an LLM-based triage engine integrated directly with the EHR via an agentic context protocol, designed to manage minor discomfort inquiries and provide safety-critical referrals. Crucially, the research establishes a multi- dimensional clinical validation framework, prioritizing the minimization of the highly detrimental under-triage rate over simple diagnostic accuracy. Finally, the system’s architecture is analyzed against the mandatory compliance obligations, including the necessity of operating as a Significant Data Fiduciary under the Digital Personal Data Protection (DPDP) Act, 2023, offering a blueprint for responsible AI deployment in regulated clinical environments