We would like, starting from the MLflow artifacts, to automatically build a Docker image. This PoC will not check scripts can run, but only generate the image based on the environment provided by the user in an experiment. The image should be the slimmest possible.