Machine learning platforms have significantly evolved since AlexNet competed in the ImageNet challenge in 2012. Here, Jordan Volz discusses the evolution of ML platforms from code- and environment-based systems to model-based systems to data-based systems. He outlines the first generation of data science tools, including frameworks for data wrangling and scientific computing, notebook environments, and collaborative data science platforms. The second generation, according to Volz, includes platforms that accelerate model development and deployment by facilitating feature engineering, model training, and model evaluation. Finally, the third generation of declarative data-centric solutions leverages feature stores, declarative AI engines, and MLOps components to automatically transform training data into AI systems.