Going from a hypothesis to a working machine learning model that infers answers in production requires a lot of time and effort. Moreover, the ability to answer questions related to specific results—such as, “what version of the code and data produced a particular inference?”—is paramount in highly regulated industries such as Financial Services. Modern development practices like continuous integration and deployment can accelerate the machine learning development process and provide a way to answer questions about data lineage. During this talk, you will learn how to combine Amazon SageMaker (a fully managed service that enables developers and data scientists to quickly and easily build, train, and deploy machine learning models at any scale) with Amazon CodeCommit, CodeBuild, and CodePipeline to create a pipeline that automatically triggers changes when either your model code or training data changes.
Presenter: Felix Candelario, Principal Global Account Solutions Architect, AWS