Let’s talk
← All work

Project · ML heart-risk model · 2026

HeartGuard

HeartWise: a 10-year heart attack and stroke risk check for Indian adults, with ML models trained on NFHS-5 data, a 3D heart you can explore, and a full model card.

HeartGuard, ML heart-risk model

What it is

HeartWise is an educational heart-health website for Indian adults. It estimates your 10-year risk of a heart attack or stroke using the WHO 2019 South-Asia risk model, and shows what its own machine-learning models learned from NFHS-5, India’s National Family Health Survey. It is clearly marked as not a medical device and not a diagnosis.

How it is built

The models are trained in Python with XGBoost, served by FastAPI, and explained in a React front end with a 3D heart you can explore and a full model card describing the data, the training and the limits.

  • A reproducible pipeline from the raw survey data to the published model.
  • A simulated 10-year Indian cohort study, with results on the Model & Data page.
  • Release checks for linting, types, tests, security, dependency advisories, speed and reproducibility.