By: on march 15, 2025

Smart Health AI

Full-stack ML system that predicts 8 diseases and supports user-driven retraining via FastAPI and MLflow

Go to project
Screenshot of Smart Health AI web interface and prediction form
~2 MIN

About Smart Health AI

Smart Health AI is a robust full-stack ML-powered application that enables users to predict the risk of eight major diseases based on simple clinical inputs. Designed with a modular FastAPI backend and a snappy React + TailwindCSS frontend, the system goes a step further by enabling user feedback loops and live retraining of models—making it an end-to-end example of real-world MLOps.

Each disease has its own prediction pipeline built using scikit-learn, with MLflow tracking model metrics and versions. Users can submit new data post-prediction, and an endpoint is available for on-demand retraining, keeping the models continually updated.

  • Predicts 8 Diseases with ML Pipelines
    Supports risk prediction for Anemia, Cardiovascular Disease, Heart Disease, Hepatitis C, Liver Disease, Lung Cancer, Stroke, and Thyroid Disease.

  • Modular ML Pipelines Per Disease
    Each disease prediction is powered by its own standalone scikit-learn pipeline, allowing easy updates, experimentation, and isolated debugging—ensuring better maintainability and control.

  • MLflow-Powered MLOps
    Model training runs, metrics, and artifacts are versioned and logged with MLflow.

  • FastAPI Backend & Docker Deployment
    Built for production with FastAPI and Docker Compose for quick orchestration of backend + frontend.

  • Frontend with React + Vite + Tailwind
    Sleek, responsive UI lets users input clinical data and view real-time predictions.


Tech Stack

  • Backend: FastAPI, Uvicorn, Pydantic
  • Frontend: React.js, TypeScript, Vite, Tailwind CSS
  • ML & MLOps: scikit-learn, Joblib, MLflow
  • Deployment & DevOps: Docker, GitHub Actions, AWS

Credits

  • MLflow: Clean and scalable experiment tracking for each disease pipeline.
  • Docker Compose: Smooth full-stack deployment in local and cloud environments.
  • FastAPI: Fast, async-ready Python backend powering predictions and feedback.
  • scikit-learn + Joblib: Traditional models, saved and reloaded efficiently.

Author

Built by Rohit Kshirsagar, Parth Lhase & Rishabh Kothari AI systems enthusiasts and founding engineers at ApexAI.
Check out more on GitHub or connect via LinkedIn.