Scaling out the serving of machine learning models in real deployments is hard - wrapping your model with a Flask API does not cut it.
