Handle Missing and Skewed Features

You are working with a supervised learning dataset that has missing values and a few heavily skewed numeric features. The model will be used in production, so the preprocessing has to be fit correctly and carried forward at inference time.