Scroll Top

MLFlow to manage the machine learning lifecycle

How do you manage your machine learning lifecycle?


The machine learning lifecycle is fundamental to products and projects that use machine learning. Whether it’s experimenting with new models, tracking the complex interplay of data, hyperparameters and performance, monitoring models in the wilds of production, or deploying crucial updates and improvements, it’s crucial that results about our ML model zoo are created, reported and propagated in a clear, consistent and actionable way.


That’s why the ML team in Henesis has integrated MLFlow into the core of our ML activities. Using MLFlow we can easily manage unlimited data, models and experiments and easily convert a successful result into a deployable and versioned model in an inference pipeline.


It has allowed us to move beyond our former mess of Jupyter notebooks and Python scripts. Instead we can frame our work in an elegant data, training, deployment and inference architecture that is framework-agnostic and easily scales, to more data and to new problems and solutions.