Best practices for operationalizing ML workloads

Many organizations start their ML journey with a few experiments of building models to solve one or more business problems. Cloud platforms, in general, and ML platforms such as SageMaker make this experimentation easy by providing seamless access to elastic compute infrastructure and built-in support for various ML frameworks and algorithms. Once these experiments have proven successful, the next natural step is to move the models into production. Typically, at this time, organizations want to move out of the research-and-development phase and into operationalizing ML.

The idea of MLOps is gaining popularity these days. MLOps, at a very high level, involves bringing together people, processes, and technology to integrate ML workloads into release management, CI/CD, and operations. Without diving into all the details of MLOps, in this section, we will discuss best practices for operationalizing ML workloads using technology. We will...