Pediatric Myopia Progression Report Generator

This project focuses on a clinical projection and reporting SaaS for ophthalmology clinics.
It turns research-based myopia progression logic into a structured, traceable, and production-ready workflow usable in real patient consultations.

💡 Project Overview

Children with myopia visit clinics regularly, but many practices lack a standardized system to translate research findings into patient-specific projections.

Questions like “How severe might it become?” and “When should intervention be considered?” often rely heavily on physician experience rather than a consistent framework.

Doctors were also spending 8 to 12 minutes per patient preparing explanations and visual summaries, without a systematic, traceable method for documenting how projections were calculated or presented.

🎯 Objective

Build a production-ready, medical-grade SaaS that:

• Projects future myopia progression using established research data
• Clearly visualizes current status and expected growth trajectory
• Generates comprehensive, stage-specific reports tailored to each patient
• Fits fast, intuitive clinical workflows
• Aligns with medical application standards and data privacy requirements

⚠️ Constraints

• Korean medical data privacy environment
• Multi-clinic tenant isolation requirements
• Infrastructure budget under $150/month
• Secure PDF generation
• Workflow completion under 2 minutes
• Design anticipating future regulatory scrutiny

🏗️ Architecture Summary

Frontend

• Next.js (TypeScript, Zod validation)
• Recharts percentile visualization

Backend

• AWS Lambda (Node.js)
• API Gateway
• PostgreSQL (RDS, private subnet)
• S3 (private report storage with signed URLs)
• CloudFront distribution
• Secrets Manager

Infrastructure

• Dev/Prod separation
• IAM isolation
• VPC-protected RDS
• CloudWatch + CloudTrail logging

🔧 Core Engineering Decisions

1. Multi-tenant isolation

• Clinic-scoped data separation via clinic_id
• Strict user-role boundaries
• No cross-clinic access paths

2. Deterministic projection engine

• Projection logic based on established axial-length percentile references
• Deterministic curve extension rules
• Stored input snapshot for each report

Each report records:

• Projection rule version
• Input parameters
• Timestamp

This prevents calculation ambiguity and supports reproducibility.

3. Server-side PDF rendering

• Implemented Lambda + Puppeteer for reliable SVG percentile rendering
• Chosen to ensure consistent clinic-grade output
• Rejected client-side rendering due to cross-browser inconsistencies

4. Cost optimization

• Initial infrastructure cost: about $180/month
• Optimized to under $120/month via NAT consolidation

🛡️ Regulatory and Audit-Defensible Design

Designed with awareness of Korean medical data privacy requirements and potential medical software audit scrutiny.

1. Deterministic, version-controlled projection engine

• Projection logic is explicitly versioned and immutable once deployed
• Every report stores projection rule version, reference dataset version, full input snapshot, and generation timestamp
• Historical reports are never recalculated retroactively

This prevents silent logic drift and enables exact reconstruction later.

2. Controlled change governance

• Projection rule updates require explicit version increments
• Documented changelog and deployment tagging (dev/prod alignment)
• Migration documentation for governed computational changes

3. Data governance and tenant boundary enforcement

• Clinic-scoped query enforcement at the service layer
• No cross-tenant data access paths
• Role-based access controls per clinic
• Encrypted storage and private object access policies
• Projection event logging with user ID and timestamp

4. Liability-aware system positioning

• Avoided diagnostic claims
• Avoided treatment recommendations
• Clearly labeled projections as research-based curve extensions

This preserves clinical utility while reducing regulatory exposure.

📈 Measurable Results

• Report generation time reduced from 1.5 minutes to 45 seconds
• AWS infrastructure cost reduced by about 30%
• Zero cross-tenant data exposure
• Consistent percentile visualization across providers

🧰 Key Skills Demonstrated

• Building regulated, compliance-aware SaaS
• Designing deterministic, version-controlled calculation engines
• Architecting multi-tenant medical data systems
• Implementing secure document generation workflows
• Operating cost-conscious AWS infrastructure

🎯 Takeaway

This project demonstrates the ability to ship production-grade systems in regulated environments with strong audit readiness and operational discipline.