An Architect's Guide to Machine Learning Operations and Required Data Infrastructure

MLOps is a set of practices and tools aimed at addressing the specific needs of engineers building models and moving them into production. Some organizations start off with a few homegrown tools that version datasets after each experiment and checkpoint models after every epoch of training. Many organizations have chosen to adopt a formal tool that has experiment tracking, collaboration features, model serving capabilities, and even pipeline features.