Assumptions

To put some boundaries to the problem, I have made the following assumptions:

• We can abstract the ML processes in three Python scripts:

  • prepare.py contains code for preparing the dataset.

  • train.py contains code for training a machine learning model.

  • evaluate.py contains code for evaluating the results of a machine learning model.

  • prepare-config.json contains the parameters to produce the Dataset (custom filters, crop, scale, flip, rotate, etc.)

  • model-config.json contains the parameters to produce the model (Algorithm, Hyper-parameters, etc.)

• We can retrieve from Looker a file raw_uuid.csv containing UUIDs selected by a complex filter and some metadata if needed.