Machine LearningHealthcareAI

Predicting the
Silent Killer

An end-to-end AI pipeline engineered to detect cardiovascular disease with 73.6%+ accuracy

68.6k
Patient Records
13
Clinical Features
0.73
Macro F1-Score
10
Models Evaluated

Our Mission

Cardiovascular disease is a leading global risk. We built this tool to act as a Digital Second Opinion. It looks at your health numbers—like blood pressure and weight—to spot patterns that might be invisible to the naked eye.

How It Works

Just like a doctor checks your vitals, our system cleans and organizes the data to ensure accuracy:

  • Smart Formatting: We convert all measurements into a standard format so nothing gets lost in translation.
  • Error Checking: We automatically filter out mistakes, like impossible blood pressure readings.
  • Fair Testing: We kept 20% of our data hidden to strictly test the AI's final performance.

Smart Analysis

The AI doesn't just guess; it calculates new health indicators to get a better picture:

BMI & Blood FlowCombined Health Signals
OptimizationVerified Accuracy: 73.6%
ArchitectureXGBoost Ensemble
Speed< 100ms Inference

Project Retrospective

A comprehensive analysis of our end-to-end journey to build a robust diagnostic AI.

1.0 Project Overview

This retrospective covers our complete lifecycle from initial data preparation to final model selection. Our primary objective was to build a robust classification model capable of accurately predicting cardiovascular disease using the cardio_cleaned_week2.csv dataset.

Goal

Systematically explore algorithms to maximize predictive performance and identify the most promising candidates.

Outcome

Successfully deployed a detailed pipeline with ~73.6% cross-validation accuracy.

CardioPredict AI

Engineered with by Jay Patel

© 2026 All Rights Reserved