This session's aim is to showcase a practical example of how to deploy and test different versions of a machine learning model in production to solve the same challenge, i.e. give a price prediction for a service.

This is achieved with Azure Kubernetes (AKS) as the containerised compute engine and Istio as a service mesh. AKS and Istio join forces and enable microservices engineers to control the way the network traffic is split across different deployed version of a machine learning model, ultimately solving challenges like A/B testing and canary releases in productionising data science.

But what if you also wanted to make sure the users of your machine learning application are always given consistent results, i.e. always been given a prediction from the same version of your model? Istio on AKS will also empower you to achieve this through sticky sessions!

The video is not available to view online.