The most comprehensive bank Interview Answer Review tooling available online.

In dev, the model training of different datasets must be tracked. This means:

* Tracking the dataset used for training, validation and testing

* Tracking ML algorithm and its parameters

* Train the algorithms + different parameters on the same training dataset, validate and test on the same validation & testing dataset.

* Compare the output of the algorithms on the test dataset.

* all of the above steps must be tracked (using MLFlow)

Then when the final algorithm is picked, the code + model must be pushed to git repo. 

The MLOps framwrok insures the code and model goes to standard workflow of dev/staging/prod. In Staging, a code review will be done (by someone else and maybe on different dataset). When all ticked, it will go to Prod and then realease