You’ve trained your first ML model and wowed the business, they want this model in production so how do you make that happen? We’ll use Flask to deploy a scikit-learn model as an API to perform inference. Then we’ll talk about challenges and next steps like how do we monitor or update our model.
There are many tutorials on how to get started with machine learning and from there it is up to you to apply techniques to your data and modeling needs. However, little time is spent on what to do if you’re successful in building your model despite more companies utilizing “full-stack” or “unicorn” models for machine learning teams which require data scientists to deploy their models to production. If you can predict with some acceptable accuracy whether a message is spam or not, how can we integrate that into the product? We’ll walk through that integration process using Flask and Docker. We’ll look at a pre-trained classification model and how we can perform inference with that model using an API. Then we’ll cover some of the limitations of this approach and how to start to think about the newest challenges in the ML engineering world like model management and monitoring.