A Data Scientists work is not done once machine learning models are in production. In this talk, Jannes will explain ways of monitoring Keras neural network models in production, how to track model decay and set up alerting using Flask, Docker and a range of self-built tools.
Once a model is in production, practitioners have to ask themselves a range of questions: - Does the live data correspond to training and testing data? - Are there anomalies and glitches, possibly exhibiting hidden problems in the model? - Is there a change in the real world that impacts our model? - Does the model (still) deliver business value? - Could we improve the model?
So far, these topics have received little attention, but as more machine learning models move into production, they become critical. In this talk, Jannes will show a range of emerging solutions from the industry, together with practical implementations using Flask to serve Keras models for computer vision.