Real-world uses

Created for this library

1. 1.

   An ML platform team keeps a frozen test set per model release so reviewers see honest, comparable evaluation.

2. 2.

   A risk modeling team uses an out-of-time test set so the production model is evaluated on data more recent than its training set.

3. 3.

   A research team writes unit tests for its training and serving code so changes do not silently break model quality.