Built from a moment that changed everything.

HeartRelief began the night my father had a heart attack.

He survived. But the days that followed were full of questions nobody could answer simply: how stressed had he been? What had his cardiovascular load looked like in the months leading up to it? Was there a pattern we could have seen?

In order to answer these questions my family had, I designed a machine learning architecture that models HRV and hemodynamic stress correlations with heart attack risk, and trained multi-layer neural networks on physiological time-series datasets. The goal was simple: an engine that outputs a cardiovascular risk score at regular intervals, so families in similar situations to mine, can actually see the trajectory instead of guessing.

HeartRelief is that engine, wrapped in something usable. A short video sample is converted into HRV and hemodynamic stress features. A structured questionnaire captures sleep, symptoms, and lifestyle context. The trained network combines them into a periodic risk score on a 0–100 scale, banded Low / Moderate / High, and plotted over time.