Scalable Healthcare Care-Management Platform
A GraphQL-driven healthcare platform with reusable architecture, strong testing, and AI-assisted workflows.
-35%
Documentation time
-45%
UI defects
+30%
New feature velocity
Overview
A care-management platform for a healthcare provider, unifying patient data, clinical workflows, and reporting behind a typed GraphQL layer with a reusable component system and an AI assistant for clinical documentation.
The business problem
Clinical teams juggled fragmented tools and slow, inconsistent screens. Data lived across services with no unified contract, and every team rebuilt the same UI primitives, creating drift, defects, and slow delivery.
The solution
I led the frontend architecture and contributed across the stack: a typed GraphQL schema as the single contract, a reusable design system, PostgreSQL-backed services, and an AI assistant that drafts clinical notes from structured encounters - always with a human in the loop.
Architecture
- Next.js App Router frontend with server components for data-heavy clinical screens.
- GraphQL gateway providing a single typed contract over multiple backend services.
- PostgreSQL with carefully designed indexes and row-level access for PHI.
- AWS delivery: S3 + CloudFront for assets, Lambda for event-driven tasks.
- AI documentation service calling OpenAI with retrieval over encounter data.
Engineering decisions
- GraphQL as the single source of truth to eliminate API contract drift.
- A versioned internal component library to enforce consistency at scale.
- Human-in-the-loop AI: drafts only, never auto-committed to the record.
Performance optimizations
- Server components + streaming to cut time-to-interactive on heavy dashboards.
- Query-level caching and pagination for large patient lists.
- Code splitting and route-level prefetching for instant navigation.
Challenges
- Protecting PHI while keeping clinician workflows fast and frictionless.
- Preventing UI drift across many squads shipping in parallel.
- Keeping AI output safe, grounded, and reviewable.
Lessons learned
- A typed contract is the highest-leverage investment in a multi-team product.
- AI earns trust through transparency and review, not autonomy.