Optimizing machine learning models is a challenging task that can greatly impact the performance of your models. Traditional methods such as grid search are often more time-consuming and can be insufficient for finding the best set of hyper-parameters. XGBoost and Optuna are libraries that allow for more efficient and effective optimization of machine learning models. In this talk, we will discuss how to use these libraries to improve your XGBoost models performance and achieve peak performance using a real life dataset that is highly imbalanced. Going beyond the simple datasets often found in tutorials. In this talk we will go through an end-to-end ML pipeline and solve a challenging scenario that predicts customer renewals based on product data usage.