The framework allows you to load and save the model in a format compatible with most popular frameworks (eg: Scikit, Pytorch, Tensorflow,…)..For a complete list have a look here.The code to save the model as an artifact is rather easy:Example of log_model call in mlFlowThe result of the fitting will be passed as the first parameter to the function, the second part is the directory..If you navigate to the UI and click on the run you will find the artifact information on the bottom of the page.Model saved on Azure from mlFlowIf you are the paranoid type you can now have a look at the blob storage account to verify that the model was actually saved.Model on blob storeNice right?.If you want you can have a look at the load feature and start building applications around the model.The full repo that was used for this demo can be found on Github.IssueThe bigger issue that I found with the framework was that all the logging was only stored in the docker container even if you have defined a storage backend..This means that if you restart the container all your logging will be lost..I logged an issue (https://github.com/mlflow/mlflow/issues/613), which got the response that the current team is redesigning the logging features..So fingers crossed.. More details