AdventHealth Research Platform

A secure, enterprise-grade research collaboration platform for cardiovascular studies, integrating Microsoft Fabric, Azure ML Studio, and Azure AI Foundry with a unified dark-theme interface.

🎯 Overview

The AdventHealth Research Platform provides a streamlined interface for data scientists and researchers to access computational resources, manage datasets, execute notebooks, deploy ML models, and collaborate on cardiovascular research projects—all while maintaining strict PHI data security and compliance.

Live Demo: https://research-portal-2fbfvjbt.devinapps.com

✨ Key Features

Unified Research Dashboard - Single interface for all research projects
Multi-Language Notebook Support - Python, R, and Scala notebooks via Microsoft Fabric
H100 GPU Inference - Real-time model inference with sub-50ms latency
PHI-Protected Datasets - Secure data access with export approval workflow
ML Model Deployment - Azure ML Studio integration with real-time metrics
Activity Tracking - Complete audit trail of all research activities
Export Approval Workflow - PI-approved data export with time-limited access
Dark Theme UI - Professional, eye-friendly interface for long research sessions

📸 Complete Demo Walkthrough

1. Dashboard - Project Overview

The main dashboard displays all active cardiovascular research projects with key metrics including notebooks, datasets, models, and recent activity.

Features shown:

Project cards with PI information
Quick stats (notebooks, datasets, models, activity count)
Direct links to Microsoft Fabric and Azure ML Studio workspaces
Status indicators and creation dates

2. Notebooks - Multi-Language Support

The platform supports Python, R, and Scala notebooks through Microsoft Fabric integration, enabling data scientists to use their preferred tools.

Features shown:

Python notebook - Real-Time Inference Pipeline with executable code
R notebook - Statistical Analysis for Heart Rate Variability
Scala notebook - Spark Streaming for ECG Data Pipeline
Language badges (Python, R, Scala)
Status indicators (Ready, Running)
"Open" button to launch in Microsoft Fabric
"Run Inference" button for executable notebooks

3. Datasets - PHI-Protected Data Access

All cardiovascular datasets are PHI-protected with security levels and access controls. Researchers can request exports through the approval workflow.

Features shown:

ECG Recordings Database (2.5M rows, 5.2 TB)
Patient Demographics (2.5M rows, 890 MB)
Medication History (6.8M rows, 1.8 GB)
Security level indicators (PHI-Protected)
Row counts and data sizes
"Request Export" button for data access
Access level badges (Full Access, Read-Only)

4. ML Models - H100 GPU Deployment

Deployed machine learning models with real-time performance metrics, GPU utilization, and accuracy tracking.

Features shown:

AFib Detection LSTM v2.3 (4 H100 GPUs, 42ms latency, 96% accuracy)
Multi-Class Arrhythmia CNN v1.8 (2 H100 GPUs, 38ms latency, 89% accuracy)
Risk Stratification XGBoost v3.0 (CPU, 12ms latency, 87% accuracy)
Real-time GPU allocation
Inference latency metrics
Model accuracy tracking
"View in ML Studio" links

5. Activity Feed - Audit Trail

Complete audit trail of all research activities including notebook executions, model training, dataset access, and export requests.

Features shown:

Chronological activity log
User attribution
Activity type indicators
Timestamp tracking
Resource identification

6. Real-Time Inference Execution

Execute inference on H100 GPUs directly from the platform with real-time results showing predictions, confidence scores, and latency metrics.

Features shown:

Real-time inference execution
ECG signal preprocessing (5000 samples)
Model predictions with confidence scores:
- Atrial Fibrillation: 78% (predicted class)
- Normal Sinus Rhythm: 12%
- Premature Ventricular Contraction: 6%
- Ventricular Tachycardia: 4%
GPU utilization (H100-0)
Inference latency: 42ms (target: <50ms)
Visual prediction bars

7. Export Requests - Approval Workflow

Track all data export requests with status, PI reviewer information, and download links for approved exports.

Features shown:

Export request tracking (EXP-2024-001, EXP-2024-002, etc.)
Status indicators (Pending, Approved, Rejected)
Dataset information and row counts
PI reviewer assignment
Request timestamps
Download links for approved exports (time-limited)
Justification tracking

8. Notifications - Real-Time Updates

System notifications for export approvals, model training completion, team member additions, and other important events.

Features shown:

Real-time notification feed
Notification types (Success, Info, Warning)
Unread indicators
Timestamp tracking
Related entity links

9. Export Request Dialog

Request data exports with justification for PI approval. The system enforces the approval workflow for all PHI-protected data.

Features shown:

Dataset selection
Row count specification
Justification text area (required)
PI reviewer assignment
Submit workflow

🏗️ Architecture

Technology Stack

Frontend:

React 18.2 with TypeScript
Tailwind CSS for dark theme styling
Vite build system
Deployed on Azure Static Web Apps

Backend:

FastAPI (Python 3.11)
Poetry for dependency management
Mock endpoints for demo (ready for real service integration)
Deployed on Fly.io

Projected Integrations (Ready to Connect):

Microsoft Fabric (workspaces, notebooks, lakehouses)
Azure ML Studio (models, compute, H100 GPUs)
Azure AI Foundry (GPT-4o, Phi-3 models)
Entra ID (B2B authentication, SSO)
Azure SQL Database (metadata, audit logs)

Network Architecture

┌─────────────────────────────────────────────────────────────┐
│                     AZURE FRONT DOOR + WAF                  │
│                   (DDoS Protection, SSL)                    │
└─────────────────────────────────────────────────────────────┘
                              ↓
┌─────────────────────────────────────────────────────────────┐
│              CUSTOM PORTAL (React SPA)                      │
│  • Dashboard UI                                             │
│  • Project listing & navigation                             │
│  • Export approval workflow UI                              │
│  • Activity feeds & notifications                           │
└─────────────────────────────────────────────────────────────┘
                              ↓
┌─────────────────────────────────────────────────────────────┐
│              BACKEND API (FastAPI)                          │
│  • REST endpoints for portal                                │
│  • Integrations with Fabric & ML Studio APIs                │
│  • Export approval workflow logic                           │
│  • Notification service                                     │
└─────────────────────────────────────────────────────────────┘
                              ↓
┌──────────────┬──────────────┬──────────────┬───────────────┐
│ Azure SQL DB │ Entra ID B2B │ Fabric API   │ Azure ML API  │
│ (Metadata)   │ (Auth)       │ (Native)     │ (Native)      │
└──────────────┴──────────────┴──────────────┴───────────────┘

🚀 Quick Start

Prerequisites

Node.js 20 LTS
Python 3.11+
Poetry (Python dependency management)

Frontend Setup

cd frontend
npm install
npm run dev

The frontend will be available at http://localhost:5173

Backend Setup

cd backend
poetry install
poetry run uvicorn app.main:app --reload

The backend API will be available at http://localhost:8000

Environment Variables

Frontend (.env):

VITE_API_URL=http://localhost:8000

Backend:

# Currently using mock data - no environment variables required
# See INTEGRATION_GUIDE.md for connecting real services

📚 Documentation

API_REFERENCE.md - Complete API documentation with all 40+ endpoints
DEMO_GUIDE.md - Detailed demo walkthrough with next steps
INTEGRATION_GUIDE.md - How to connect real Azure services
DEPLOYMENT_GUIDE.md - Production deployment instructions
SECURITY_IMPLEMENTATION.md - Security features and compliance
SECURE_ARCHITECTURE.md - Network security and architecture

🔐 Security Features

Entra ID B2B Authentication - SSO with MFA support
PHI Data Protection - All datasets marked as PHI-protected
Export Approval Workflow - PI approval required for data exports
Time-Limited Access - Export downloads expire after 24 hours
Audit Logging - Complete activity trail for compliance
Network Isolation - VNet with no internet gateway (production)
Private Endpoints - Secure connectivity to Azure services

🔌 API Endpoints

Core Endpoints

GET  /api/v1/user/profile              - Get current user profile
GET  /api/v1/user/projects             - Get user's projects
GET  /api/v1/projects/{id}             - Get project details
GET  /api/v1/projects/{id}/notebooks   - Get project notebooks
GET  /api/v1/projects/{id}/datasets    - Get project datasets
GET  /api/v1/projects/{id}/models      - Get deployed models
POST /api/v1/exports/request           - Submit export request
GET  /api/v1/exports/my-requests       - Get export requests
POST /api/v1/notebooks/{id}/run-inference - Execute inference

Integration Endpoints (Projected)

# Microsoft Fabric
GET  /api/v1/fabric/workspaces
GET  /api/v1/fabric/workspaces/{id}/lakehouses
POST /api/v1/fabric/notebooks/{id}/execute

# Azure ML Studio
GET  /api/v1/mlstudio/compute
GET  /api/v1/mlstudio/endpoints
POST /api/v1/mlstudio/endpoints/{id}/invoke

# Azure AI Foundry
GET  /api/v1/foundry/models
POST /api/v1/foundry/models/{id}/deploy

# Entra ID
POST /api/v1/auth/login
POST /api/v1/auth/refresh
GET  /api/v1/auth/validate

See API_REFERENCE.md for complete documentation.

📊 Mock Data

The platform currently uses mock cardiovascular research data for demonstration:

Projects:

Cardiac Arrhythmia Prediction Study
Heart Failure Readmission Analysis
Coronary Artery Disease Risk Modeling
Real-Time ECG Inference on H100

Datasets:

ECG Recordings Database (2.5M rows, 5.2 TB)
Patient Demographics (2.5M rows, 890 MB)
Medication History (6.8M rows, 1.8 GB)

Models:

AFib Detection LSTM v2.3 (96% accuracy, 4 H100 GPUs)
Multi-Class Arrhythmia CNN v1.8 (89% accuracy, 2 H100 GPUs)
Risk Stratification XGBoost v3.0 (87% accuracy, CPU)

🔄 Next Steps - Connecting Real Services

To connect real Azure services, follow these steps:

1. Microsoft Fabric Integration

# backend/app/services/fabric.py
from azure.identity import DefaultAzureCredential

credential = DefaultAzureCredential()
token = await credential.get_token("https://api.fabric.microsoft.com/.default")

# Replace mock_notebooks with real Fabric API calls
response = await httpx.get(
    f"https://api.fabric.microsoft.com/v1/workspaces/{workspace_id}/notebooks",
    headers={"Authorization": f"Bearer {token.token}"}
)

2. Azure ML Studio Integration

# backend/app/services/azureml.py
from azure.ai.ml import MLClient
from azure.identity import DefaultAzureCredential

ml_client = MLClient(
    credential=DefaultAzureCredential(),
    subscription_id="your-subscription-id",
    resource_group_name="your-resource-group",
    workspace_name="your-ml-workspace"
)

# Get real deployed models
endpoints = ml_client.online_endpoints.list()

3. Entra ID Authentication

// frontend/src/config/msalConfig.ts
import { Configuration } from '@azure/msal-browser';

export const msalConfig: Configuration = {
  auth: {
    clientId: process.env.VITE_CLIENT_ID,
    authority: 'https://login.microsoftonline.com/your-tenant-id',
    redirectUri: 'https://research.adventhealth.com',
  }
};

See INTEGRATION_GUIDE.md for complete integration instructions.

🧪 Testing

Backend Tests

cd backend
poetry run pytest

Frontend Tests

cd frontend
npm run test

API Testing

Use the provided Postman collection in API_REFERENCE.md or test with curl:

# Get user projects
curl http://localhost:8000/api/v1/user/projects

# Get project notebooks
curl http://localhost:8000/api/v1/projects/proj-004/notebooks

# Run inference
curl -X POST http://localhost:8000/api/v1/notebooks/nb-006/run-inference

📦 Deployment

Frontend Deployment (Azure Static Web Apps)

cd frontend
npm run build
# Deploy dist/ folder to Azure Static Web Apps

Backend Deployment (Fly.io)

cd backend
fly deploy

See DEPLOYMENT_GUIDE.md for detailed deployment instructions.

🤝 Contributing

This is a private research platform for AdventHealth. For questions or support, contact:

Project Lead: Dr. Sarah Smith (dr.smith@adventhealth.com)
Technical Support: research-platform@adventhealth.com

📄 License

🔗 Links

Live Demo: https://research-portal-2fbfvjbt.devinapps.com
Backend API: https://app-tiouegnz.fly.dev
API Documentation: https://app-tiouegnz.fly.dev/docs
Devin Session: https://app.devin.ai/sessions/2e347798f7e842269122e54741b5d338

Built with ❤️ for cardiovascular research at AdventHealth

Name		Name	Last commit message	Last commit date
backend		backend
frontend		frontend
infrastructure		infrastructure
screenshots		screenshots
API_ENDPOINTS.md		API_ENDPOINTS.md
API_REFERENCE.md		API_REFERENCE.md
DEMO_GUIDE.md		DEMO_GUIDE.md
DEPLOYMENT_GUIDE.md		DEPLOYMENT_GUIDE.md
IMPLEMENTATION_SUMMARY.md		IMPLEMENTATION_SUMMARY.md
INTEGRATION_GUIDE.md		INTEGRATION_GUIDE.md
README.md		README.md
SECURE_ARCHITECTURE.md		SECURE_ARCHITECTURE.md
SECURITY_CHECKLIST.md		SECURITY_CHECKLIST.md
SECURITY_IMPLEMENTATION.md		SECURITY_IMPLEMENTATION.md
SECURITY_REVIEW.md		SECURITY_REVIEW.md

Latest commit

History

Folders and files

Repository files navigation

AdventHealth Research Platform

🎯 Overview

✨ Key Features

📸 Complete Demo Walkthrough

1. Dashboard - Project Overview

2. Notebooks - Multi-Language Support

3. Datasets - PHI-Protected Data Access

4. ML Models - H100 GPU Deployment

5. Activity Feed - Audit Trail

6. Real-Time Inference Execution

7. Export Requests - Approval Workflow

8. Notifications - Real-Time Updates

9. Export Request Dialog

🏗️ Architecture

Technology Stack

Network Architecture

🚀 Quick Start

Prerequisites

Frontend Setup

Backend Setup

Environment Variables

📚 Documentation

🔐 Security Features

🔌 API Endpoints

Core Endpoints

Integration Endpoints (Projected)

📊 Mock Data

🔄 Next Steps - Connecting Real Services

1. Microsoft Fabric Integration

2. Azure ML Studio Integration

3. Entra ID Authentication

🧪 Testing

Backend Tests

Frontend Tests

API Testing

📦 Deployment

Frontend Deployment (Azure Static Web Apps)

Backend Deployment (Fly.io)

🤝 Contributing

📄 License

🔗 Links

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages