Selected Projects
Four independent builds — from convolutional networks to a real-time wearable pipeline.
ECG Stress Monitor
A real-time stress-monitoring system that streams ECG from a Polar H10 chest strap, extracts HRV features, and estimates stress with a LightGBM model — synced with Google Calendar for context on what you were doing.
View on GitHubSmoking Status Prediction
A leakage-safe 5-model ensemble (LightGBM ×2, XGBoost, TabNet, ResMLP) with stratified K-fold CV and out-of-fold evaluation — ranked #1 private / #3 public LB (AUC 0.883).
View on GitHub
Brain Tumor Detection with CNN
A convolutional neural network trained on MRI scans to classify tumor presence — my first deep dive into CNN architecture, image preprocessing, and hyperparameter tuning.
View on GitHub
Flower Recognition App
A Django web app that classifies 17 flower species from user-uploaded photos — image upload pipeline, CNN inference routing, and dynamic results rendering, built end to end.
View on GitHub