Aethera Health developed AetheraOS, an EHR and practice management platform for integrative
and naturopathic medical clinics handling dispensary tracking and AI clinical charting.
Having validated their system on-premises in Docker Compose, the team lacked the production
cloud engineering required to launch their commercial cohort of 25 medical practices. The
deployment needed to meet strict HIPAA Security Rule technical safeguards, guarantee
multi-tenant clinical data isolation, and provide low-latency LLM inference without
incurring heavy GPU server debt.
The Solution & Technical Highlights
Serverless Multi-AZ Compute
Architected a three-tier Virtual Private Cloud (VPC) across multiple Availability
Zones on AWS ECS Fargate, abstracting container OS patching and auto-scaling tasks
behind an Application Load Balancer and AWS WAF.
Unified Relational & Vector Store
Provisioned Multi-AZ Amazon RDS PostgreSQL 16 with the pgvector extension and
Row-Level Security (RLS), consolidating relational clinic data and semantic
botanical monographs into one ACID engine with strict tenant isolation.
Private Asynchronous GenAI Pipeline
Engineered an asynchronous RAG queue via Amazon SQS and ECS worker tasks querying
AWS Bedrock (Claude 3.5) across AWS PrivateLink VPC Endpoints, ensuring clinical
prompts never traverse public internet fiber.
HIPAA Compliance & WORM Audit Controls
Implemented AES-256 KMS encryption at rest, TLS 1.3 in transit, and continuous WAL
backups (RPO < 5 min, RTO < 30 min) with pgaudit logs archived to Amazon S3
Object Lock in Compliance Mode for mandatory 6-year immutability.
Key Outcomes & Impact
Achieved 100% reproducible Infrastructure as Code (IaC) with automated GitHub
Actions CI/CD vulnerability scanning via Trivy, delivering production cutover on
schedule within 8 weeks.
Reduced baseline hosting spend to ~$356/month across 25 pilot clinics (~$14.25 per
practice), eliminating over $700/month in idle GPU compute while operating under a
signed AWS Business Associate Agreement (BAA).
