What strategies would you use to handle imbalanced datasets when deploying machine learning models?

What strategies would you use to handle imbalanced datasets when deploying machine learning models?