Isn't "keras for quick experiments" misleading nowadays?

[andrey-klochkov-liftoff]

The project README says the following:

Provide tf.keras.Model like interface for quick experiment . example
Provide tensorflow estimator interface for large scale data and distributed training . example

Is this still actual? Isn't tensorflow estimator API somewhat deprecated, and it's not superior to Keras for large scale and distributed learning?