BridgeFlow AI

Modern AI that bridges human requests with legacy healthcare systems.

Comment

BridgeFlow AI

BridgeFlow AI is an agentic healthcare workflow assistant that automates patient registration in legacy healthcare systems using AI.

The system allows users to describe a patient registration request in natural language. BridgeFlow AI then extracts patient information, validates the data, detects duplicates, and safely automates the registration process.

This project was built for the Amazon Nova AI Hackathon.

Problem

Many healthcare organizations still rely on legacy portals that require manual patient registration. These systems create several problems: - Slow manual workflows - Human data entry errors - Duplicate patient records - Poor data validation - Inefficient administrative processes Healthcare providers need automation that improves efficiency while maintaining safety and accuracy.

Solution

BridgeFlow AI introduces an AI-powered workflow layer between users and legacy healthcare systems. Instead of manually filling out forms, users can simply type a request like: Register patient John Doe born 1999-05-10 insured Aetna phone 1234567890

BridgeFlow AI automatically: 1. Understands the request using AI 2. Extracts structured patient information 3. Validates healthcare data 4. Calculates a confidence score 5. Detects duplicate patients 6. Automates legacy system registration If the system detects a potential risk (such as incorrect data or missing information), it switches to a human-in-the-loop workflow to ensure safe submission.

Key Features

  • Natural language patient registration
  • AI-powered request parsing
  • Risk scoring and validation layer
  • Duplicate patient detection
  • Legacy system automation
  • Patient registry database
  • Human-in-the-loop correction workflow

Architecture

User Request ↓ AI Parser (Amazon Nova) ↓ Data Validation Layer ↓ Risk &Confidence Scoring ↓ Duplicate Detection ↓ Automation Agent ↓ Legacy Healthcare Portal ↓ Patient Registry Database

Example Scenarios

Normal Registration

Register patient Emily Carter born 1988-11-22 insured Blue Cross phone 7135551234 Result: - AI extracts patient data - Validation passes - Agent submits registration - Patient is stored in the database

Duplicate Patient Detection

Register patient John Doe born 1999-05-10 insured Aetna phone 1234567890 Result: - Existing patient record detected - System prevents duplicate registration - Warning message displayed

Risk Detection

Register patient David Kim born 1992-08-30 insured UnitedHealth phone 12345 Result: - Phone number fails validation - Risk level flagged - Legacy form opens for manual verification

Tech Stack

  • Python
  • Flask
  • SQLite
  • Amazon Nova AI
  • Automation agent workflow

Project Structure

bridgeflow-ai │ ├─ agent │ ├─ agent_api.py │ ├─ agent.py │ ├─ ai_parser.py │ ├─ db.py │ ├─ init_db.py │ ├─ legacy_agent.py │ ├─ requirements.txt │ └─ test_bedrock.py │ ├─ dashboard │ ├─ public │ ├─ src │ ├─ package.json │ └─ package-lock.json │ ├─ docs ├─ legacy-portal ├─ .gitignore └─ README.md

Running the Project

  • Backend cd agent pip install -r requirements.txt python agent_api.py

The backend server runs at: http://127.0.0.1:5000

  • Frontend cd dashboard npm install npm start

Future Improvements

  • Voice-based patient intake
  • Insurance verification automation
  • Appointment scheduling workflows
  • Integration with real healthcare APIs

Hackathon:

This project was built for the Amazon Nova AI Hackathon. BridgeFlow AI demonstrates how AI agents can safely automate real-world healthcare workflows while maintaining human oversight.

Built With

Share this project:

Updates