Mohammad Haroon Portfolio · Case study
MedMatch AI
Heart Disease Risk Prediction System
AI-powered healthcare app that predicts heart disease risk and generates personalized recommendations.
Overview
MedMatch AI is an educational heart disease risk assessment app. Users submit clinical features in a React form; a FastAPI backend scores risk with a Random Forest model and returns urgency, doctor advice, and lifestyle tips from rule-based logic. It demonstrates ML serving end to end and is not a medical device for real clinical decisions.
About the project
An AI-powered healthcare demo that analyzes patient health indicators to predict heart disease risk and generate personalized recommendations. The Random Forest pipeline and FastAPI inference layer report strong held-out performance (about 81.5% accuracy and 0.914 ROC-AUC in project documentation). Educational use only — not for clinical diagnosis.
Role & contribution
As Full Stack & ML Developer, Mohammad Haroon built the React intake UI, FastAPI inference API, Random Forest training pipeline, and rule-based recommendation layer.
Frontend & backend
Frontend technologies
Backend technologies
Challenges & solutions
Feature alignment at inference
ProblemOne-hot encoded columns must match the exact training schema or predictions break.
SolutionPersist feature_columns.pkl and align request features before scaling and scoring.
Explainable recommendations
ProblemA raw probability alone does not tell a user what to do next.
SolutionKeep recommendations rule-based so each tip maps to concrete inputs such as cholesterol or blood pressure thresholds.
Local full-stack DX
ProblemVite and FastAPI run on different ports and need safe cross-origin calls.
SolutionCORS-enabled FastAPI endpoints for local Vite development with /health and /predict.
Educational safety
ProblemHealth demos can be mistaken for clinical tools.
SolutionShip an explicit disclaimer that MedMatch AI is educational and not a medical device.
Project details
- Patient health indicator intake form
- Random Forest risk prediction API
- Urgency, doctor advice, and lifestyle tips
- Train/export/load model artifact workflow
- 81.5% accuracy and 0.914 ROC-AUC (project-reported)
- Clear educational / non-clinical disclaimer
Full technology stack
Frequently asked questions
What is MedMatch AI?
MedMatch AI is an educational React and FastAPI app that scores heart disease risk with a Random Forest model and returns urgency plus lifestyle guidance.
Is MedMatch AI a medical device?
No. MedMatch AI is an educational demo by Mohammad Haroon and must not be used for real clinical decisions.
What stack does MedMatch AI use?
React, Vite, Tailwind CSS, FastAPI, scikit-learn Random Forest, and joblib model artifacts trained on UCI/Kaggle heart-disease data.