Leonardo Alvarino

Machine Learning & Healthcare

Early Sepsis Detection in the ICU Using Machine Learning

Built an end-to-end ML pipeline on 790,000+ ICU patient records to detect sepsis a median of 23 hours before clinical diagnosis. Achieved AUC of 0.885 using XGBoost with SHAP explainability.

Data Wrangling & Visualization

Analyzing Value Retention in the BMW Used Car Market

Researched which BMW I should buy next — balancing fun and savings — by analyzing two Kaggle datasets on used car sales. Used linear regression to model depreciation across engine sizes and SUV models, revealing the 2 Series (3.0L) and X4 as the best bets for value retention.

Data Wrangling & ML

Predicting Tomorrow's Closing Price of the S&P 500

Built a Random Forest model to predict next-day S&P 500 direction using Yahoo Finance and FRED API data with feature engineering across 5 time horizons. Backtested a model-guided dollar-cost averaging strategy against passive DCA, both returning ~15% on $27.5K invested.

Machine Learning

Predicting House Age using Machine Learning

Trained a Random Forest classifier to predict whether a house was built before or after 1980 using 48 features from a 22,900-row dataset. Achieved 92.3% accuracy, with living area, number of bathrooms, and stories as the top predictors.

Data Wrangling & ML

Market Stress Detection

Built a market stress early warning system using S&P 500 data. Engineered risk features — rolling volatility, cumulative returns, and drawdowns — then trained a Logistic Regression model to estimate the probability of market stress. Found that risk spikes days before price drops, showing markets can look calm while becoming fragile.