AI Triage & Intake System for a Precision Wellness Platform
Architected a secure full-stack wellness platform with a LangGraph-powered AI triage agent, RAG-backed health coaching, and real-time Socket.io handoffs between patients and human coaches.
Tech Stack
Project Overview
The client operates in the preventative healthcare and precision wellness space, empowering individuals through biomarker testing, personalized supplement plans, and human health coaching. I was tasked with architecting and developing a secure, full-stack application focused on streamlining patient intake and engineering an advanced AI-driven triage and coaching system that bridges automated insights and human medical professionals.
The Challenge
- Challenge 01
Building a secure environment capable of handling sensitive user health metrics and lifestyle data.
- Challenge 02
Moving beyond basic chatbot functionality to create a stateful, context-aware AI agent for complex, multi-step triage conversations.
- Challenge 03
Enabling real-time, seamless communication handoffs between the AI system, patients, and human coaching staff.
Technical Implementation
Intelligent Patient Triage via LangGraph
Instead of relying on rigid decision trees, I engineered a conversational AI agent using LangChain and LangGraph. LangGraph was crucial for managing complex, non-linear state flows during patient intake, allowing the AI to autonomously perform initial triage, gather lifestyle and symptom context, and answer health questions using RAG backed by scientific data.
Dynamic Routing & Real-Time Collaboration
The AI agent was designed to understand the limits of its medical scope. Using intelligent routing logic, the system dynamically escalates patients to the correct human health coach when necessary. Socket.io powers real-time websocket connections, allowing coaches to instantly review AI-gathered context and jump into live sessions.
Secure Full-Stack Architecture
Developed the frontend using React and Next.js for patients and administrative coaching staff. The Node.js and PostgreSQL backend on AWS ensured all data—from questionnaires to aggregated biomarker results—was stored and managed efficiently as the user base scaled.
Key Achievements & Impact
- Impact 01
The AI-powered triage agent drastically reduced manual administrative workload for human coaches by gathering preliminary health context and filtering routine inquiries.
- Impact 02
Patients received immediate, context-aware responses and guidance, significantly improving onboarding and engagement with preventative health plans.
- Impact 03
LangGraph state management and robust AWS infrastructure provided a highly scalable foundation for AI-assisted, human-centric wellness coaching.