Steppy is a lightweight, open-source Python library for fast and reproducible machine learning experimentation. It provides a simple programming model using Step and Transformer objects to represent computational steps and operations on data. Steppy handles model training and testing, saves intermediate results and trained models, visualizes the graph of computations, and makes it easy to exchange steps. It aims to help data scientists and machine learning researchers reduce technical debt and more efficiently run experiments for their projects. Users can get started with Steppy either from scratch or by reviewing examples on GitHub to help transition their work to the Steppy framework.