Clinical AI Platform
Core Engineer — Kinwits, for a US healthcare provider
Jul 2025 – Present
A HIPAA-compliant AI platform in production that generates treatment plans from EHR, CRM, and lab data.
EHR
Cerbo
CRM
GoHighLevel
Lab Results
AI Pipeline
LangGraph · Claude on Bedrock
Treatment Plan
Generated draft
Provider Review
Approve, edit, or flag
The Platform
The platform generates clinical treatment plans by combining data from an EHR (Cerbo), a CRM (GoHighLevel), and lab results, then running it through an LLM pipeline with a provider-facing review and feedback workflow.
AI Provider Architecture
Migrated the platform's LLM integration from OpenAI to Anthropic Claude across the codebase, behind a provider-abstraction layer (text, JSON, vision, and structured-output APIs) that enables configuration-driven switching and safe model upgrades.
RAG Recommendation Engine
Architected the supplement-recommendation engine: RAG over AWS Bedrock Knowledge Bases with deterministic pre-LLM selection across a 500+ product catalog. It's the platform's largest subsystem.
Human-in-the-Loop Feedback
Designed an asynchronous clinician-feedback pipeline that classifies provider corrections to AI-generated plans and proposes fixes, closing the human-in-the-loop quality cycle.
Reliability & Compliance
Improved reliability and performance: redesigned lab-result retrieval to read stored results before LLM extraction, sped up PDF processing ~17x, and cut CRM API calls ~90% through caching.
Hardened HIPAA compliance across the stack: PHI scrubbing in logs, customer-managed encryption, immutable audit archives, and LLM calls routed within the compliance boundary. Rebuilt the backend CI/CD pipeline with required tests, vulnerability scanning and smoke tests.
Provider-Facing Web App
Built the provider-facing web app in Next.js: a feedback dashboard, AI-insights analytics, scheduled-plan review with one-click EMR write-back, and a rich-text treatment-plan editor with version history.