Pandas: refactoring data cleaning pipeline to work with pipes and functions

How to move from traditional 'R' like style of pandas to a 'flow' like style, where you pipe the dataframe trough data manipulation stages. This gives a better readability and modularity to the pipeline.