NovaCare: AI-Powered Healthcare Workflow Automation using Amazon Nova

🌟 Project Overview

NovaCare is an end-to-end healthcare automation system built to streamline patient triage, voice registration, document processing, and UI automation for hospital booking systems. Powered by Amazon Nova models—including Nova 2 Sonic for voice, Nova 2 Lite for reasoning, Multimodal Embeddings for medical documents, and Nova Act for UI interactions.

🚀 Features

• Voice AI Interface (Nova 2 Sonic): Captures patient audio, transcribes it, and converts it to a structured JSON intent.
• Medical Document Understanding: Extracts entities from multimodal prescription uploads using Nova Multimodal Embeddings.
• Reasoning Engine (Nova 2 Lite): Triages symptoms and provides actionable healthcare directions.
• UI Automation (Nova Act): Automates interacting with web portals to book appointments.

📊 Architecture Diagram

graph TD
    A[Patient Voice / Form] -->|Audio| B(Nova 2 Sonic API)
    B -->|Transcribed Intent| C{Backend FastAPI}
    D[Prescription Docs PDF/Image] -->|Upload| E(Nova Multimodal Embeddings)
    E -->|Structured Medical Record| C
    C -->|Patient Data / Symptoms| F(Nova 2 Lite Reasoning Engine)
    F -->|Triage & Next Steps| C
    C -->|Trigger Action| G(Nova Act Agent)
    G -->|UI Automation| H[Hospital/Insurance Portal]
    C --> I[(PostgreSQL Database)]

🗄️ Database Schema

We use PostgreSQL.

• patients: id, name, dob, contact, created_at
• documents: id, patient_id, file_url, parsed_data (JSON), uploaded_at
• visits: id, patient_id, triage_result, assigned_department, status
• agents_logs: id, action, status, timestamp

🛠️ Tech Stack

• Frontend: React, Vite, Tailwind CSS
• Backend: Python FastAPI, SQLAlchemy
• Database: PostgreSQL
• AI Models: Amazon Nova (2 Sonic, 2 Lite, Act, Embeddings)

🏃 Setup Instructions (Local & AWS)

Prerequisites

• Python 3.9+
• Node.js 18+
• PostgreSQL
• AWS Account with Nova models access.

Local Setup

1. Database: Create a Postgres database named novacare.
2. Backend: bash cd backend python -m venv venv source venv/bin/activate pip install -r requirements.txt uvicorn main:app --reload
3. Frontend: bash cd frontend npm install npm run dev

AWS Deployment

1. Package the FastAPI backend using Docker / AWS Lambda Web Adapter.
2. Deploy the Frontend to AWS S3 + CloudFront.
3. Use Amazon RDS for PostgreSQL.
4. Set up IAM roles with exact permissions for Amazon Nova model invocations.

Built With

• amazon-web-services
• node.js
• nova
• python
• tailwind