This paper investigates the application of the XGBoost model, combined with Bayesian hyperparameter tuning, for optimizing the quality of brewed coffee flavour. XGBoost, a powerful gradient boosting algorithm, is used to model the complex relationships between brewing parameters and flavour characteristics. To enhance model performance, Bayesian optimization is employed to systematically tune critical hyperparameters such as learning rate, maximum tree depth, and number of estimators. This approach allows for efficient exploration of the hyperparameter space, balancing model complexity and accuracy. The optimized model not only improves the prediction of coffee flavour quality but also provides insights into the key factors influencing flavour, such as temperature, and bean characteristics. This method demonstrates the potential for advanced machine learning techniques to refine and enhance the consistency and overall quality of brewed coffee, offering practical implications for the coffee.