Smart Bed AI

Are we optimizing hospital resources with intelligence? Yes!

Hackathon Context

This project was built during a hackathon focused on healthcare optimization and real-time resource allocation. The goal was to demonstrate how AI can improve hospital bed management and patient forecasting in a fast-paced environment.

Project Overview

Smart Bed AI is a system designed to forecast patient admissions, allocate available beds using optimization techniques, and provide a web-based dashboard for hospital staff to manage wards, simulate patient flow, and react to changing conditions.

Live Demo

A live demo can be accessed at smart-bed-ai (replace with real URL when deployed). The demo allows users to log in, view current bed allocations, forecasted demand, and run simulations.

How It Works

1. Forecasting: Historical admission data is fed into a forecaster (likely using ML or time-series models) located in backend/engine/forecaster.py.
2. Allocation: A Mixed Integer Linear Program (MILP) in backend/engine/milp_allocator.py determines optimal bed assignments given predicted demand and ward constraints.
3. Backend: FastAPI-based service defined in backend/main.py with routers handling allocation, forecasting, patients, wards, and more.
4. Frontend: React with Vite (src/) presents dashboards, controls, and simulation views.
5. WebSocket: Real-time updates are pushed via WebSocket to client dashboards.

Tech Stack

• Backend: Python, FastAPI, Pydantic, SQLAlchemy (presumably), uvicorn
• Optimization: MILP solver (could be OR-Tools or similar)
• Frontend: TypeScript, React, Vite, TailwindCSS or custom UI components
• Database: SQLite/PostgreSQL (check backend/database.py for config)
• Testing: Pytest for backend (test_beds_api.py, others); likely Jest for frontend

Setup Instructions

1. Clone repo: git clone <repo-url>
2. Navigate to backend: cd backend and create a virtualenv: python -m venv venv then source venv/bin/activate.
3. Install dependencies: pip install -r requirements.txt.
4. Initialize database: python seed.py or run migration scripts.
5. Start backend server: uvicorn main:app --reload.
6. Open another terminal for frontend: cd ../src and run npm install && npm run dev.
7. Visit http://localhost:3000 (or appropriate port) to access the UI.

Troubleshooting

• Cannot connect to database: Verify DATABASE_URL in environment or backend/database.py settings.
• Port conflicts: Ensure backend and frontend ports (8000 and 3000) are free.
• Dependency issues: Run pip install --upgrade -r requirements.txt and npm audit fix.

Demo Mode

The application supports a demo mode where sample data is preloaded and user actions are sandboxed. Enable by setting DEMO_MODE=true in the environment before starting the backend.

Demo Controls

Use the web UI to:

• Start/stop patient simulation
• Adjust forecast parameters
• Reset bed allocations
• View historical trends

Project Structure

backend/             # Python API and logic
  engine/            # Forecasting and allocation engines
  routers/           # FastAPI route handlers
  scripts/           # utility scripts
  models.py          # data models
  database.py        # DB configuration
  main.py            # FastAPI application
src/                 # Frontend React application
  components/        # UI components and views
  lib/               # shared utilities
index.html           # main HTML entry
package.json         # frontend dependencies

Environment Variables

• DATABASE_URL - connection string for the database
• DEMO_MODE - enable demo data and sandbox environment
• SECRET_KEY - application secret for signing tokens
• PORT - backend server port
• FRONTEND_URL - allowed origins for CORS

Running Tests

• Backend: pytest from the project root. See test_beds_api.py and test_quick_add.py.
• Frontend: npm run test inside src if tests exist.

Performance SLAs

• Forecast generation: <500ms per run
• Allocation solver: <1s for typical ward sizes
• API response: <200ms for read operations

Hard Constraints

• Each ward has a hard bed count limit
• Patients cannot be double-assigned
• Forecasting models must use only past data

Use Case Story

A nurse logs into the dashboard on a busy morning. She views the forecast, which predicts a surge of admissions in the next 24 hours. Using the simulator, she tests reallocating beds from less-critical wards. The MILP engine suggests an optimal layout, and she approves it. Real-time updates push to all staff screens as patients are admitted, keeping everyone informed.

Team

• M Tanusree Reddy
• Nihal DR
• P Devesh Reddy

Thank you for checking out Smart Bed AI! Enhance hospital care with intelligent resource management.