Data-driven and machine learning-based equipment monitoring solutions have become increasingly sophisticated and prevalent. Continuous cloud-based execution of trained workflows is now a key requirement of gaining operational value from those solutions.

This talk is aimed at data scientists and machine learning engineers developing machine learning-based workflows for equipment monitoring in heavy asset and other industries. Specifically this will be of interest to third party developers like ourselves who are interested in learning how off the shelf offerings for deploying such workflows as web-hosted APIs compare between the major cloud providers.

We will present some example workflows developed using open source python packages including pandas, scikit-learn, Keras and ADTK (anomaly detection toolkit). We will then compare the frameworks for deploying those workflows as cloud-based web services using Google Cloud Platform, Microsoft Azure and Amazon Web Services. In particular, we will present comparisons of the wrapper code required and our findings on any current framework limitations in moving these offline trained workflows to the cloud.