Many times, during your working career as a data scientist, you'll find yourself having need of a predictor decoupled from the code you're currently working on; for example, as follows:
- You're developing an app for your phone, and you want to save on memory
- You're coding in a non-Python programming language (Java, Scala, C, C++, and so on) and you need to call the predictor you've developed in Python
- You're operating on big data, and the model is trained in the same remote location where the data is stored
In all these cases, it would be nice to have a service over HTTP that does predictions-as-a-service, or generically, any machine-learning-as-a-service (ML-AAS).
Bottle, a ...
