When you start a data science project within your team, you start by sharing notebooks, code prototypes and finally developing production-ready code to train, evaluate and deploy your machine learning models. However, if more than one person is running simulations with hyper-parameter optimisations, you need a centralised way of recording everything about your experiments: code, data, config and results. This is the only way your results can be reproducible across many dimensions: between colleagues, at different points in time, etc.
Here is where MLflow comes into play. MLflow is an open source platform to manage the machine learning lifecycle.