Kedro is an open-source Python framework for creating reproducible, maintainable and modular data science code. It borrows concepts from software engineering best-practice and applies them to machine-learning code; applied concepts include modularity, separation of concerns and versioning.