
Hi, I'm Dion
I'm an MSc Artificial Intelligence (Biomedicine and Healthcare) student at University College London with a strong interest in data science and machine learning.
I first got into AI through Kaggle, where I started experimenting with beginner competitions. That experience got me hooked on solving real-world problems with data and led me to pursue this degree.
MY PORTFOLIO
These projects showcase my practical experience in machine learning and AI, with a focus on healthcare, tabular modeling, and interpretable solutions.

House Prices Prediction (Kaggle “Advanced Regression Techniques”) Python, sklearn, XGBoost, LightGBM
Developed a regression pipeline to predict house prices using 79 features and models like XGBoost and LightGBM, ranking in the top 1% on Kaggle’s leaderboard. Used feature engineering, imputation, and ensemble modeling to achieve a private RMSLE of 0.1172.

Patient Readmission Prediction (Diabetes Dataset)
Python, pandas, scikit-learn, XGBoost, imbalanced-learn
Built predictive models for hospital readmission using a 101K-record diabetes dataset with 50 features, applying encoding strategies, standardization, and hyperparameter tuning. Achieved F1 = 0.590 on test data with a Random Forest model after reducing to 25 key features via importance-based selection.

Heart Disease Prediction with CatBoost
Python, pandas, scikitlearn, CatBoost, Matplotlib
Built a heart disease classifier using CatBoost on a 918-patient clinical dataset, applying encoding, scaling, and 5-fold CV to reach 91% test accuracy (AUC = 0.91). Outperformed baselines (Logistic Regression, Random Forest) and identified key predictors (cholesterol, ST-slope, oldpeak) consistent with clinical risk factors.