HIPAA-Compliant Healthcare AI & Full Stack Engineering: Integrating LLMs, Vision OCR & Medplum FHIR

Deploying generative AI and full stack patient portals in healthcare requires strict compliance with HIPAA privacy/security rules, Business Associate Agreements (BAAs), and HL7 FHIR R4 clinical data interoperability.

As a Senior Freelance Healthcare & AI Programmer (Hire Aldian Fazrihady on Upwork), I have designed and delivered end-to-end clinical portals, AWS Bedrock Vision insurance card extractors, BoldSign cryptographic e-signature workflows, and Azure Communication Services telehealth rooms.

1. Zero-Data-Retention BAA Inference with AWS Bedrock & Vertex AI

• Encrypted Binary Handling: Patient insurance cards and clinical PDFs are stored as encrypted Medplum Binary resources in S3 with KMS customer-managed keys.
• Structured FHIR Resource Mapping: Multimodal Vision LLMs extract subscriber IDs, group numbers, and payer names into validated FHIR Coverage, Patient, and QuestionnaireResponse resources.

2. Multi-Tenant Identity & SMART-on-FHIR Scoping

• Multi-Tenant HIPAA Identity with AWS Cognito, OAuth 2.0, and SMART-on-FHIR
• Vision AI in Healthcare: Building an Insurance Card Extractor with AWS Bedrock and Medplum Binaries
• Scaling the Medplum Search Index: Mitigating API Search Latency in Large Databases