The growth of machine learning (ML) has led to an explosion of new opportunities, but also introduced challenges in productionizing ML models. Historically, data analytics teams were tasked with building ML models for internal use cases, while Silicon Valley companies were pushing the boundaries of using ML in customer-facing products. The early days of production ML were marked by difficulties in working with data, including discovering sources, building reliable pipelines, and managing training-serving skew. However, advancements in DevOps tools and MLOps have made it easier for businesses to launch ML models, with a growing demand for production-readiness and real-time predictions. Modern MLOps ecosystems are becoming less complicated, with trends such as collapsing data storage costs, the modern data stack, and embedded teams making it easier for data scientists and engineers to build high-quality models. The future of ML will be marked by advancements in managing the data lifecycle for machine learning, including a powerful ML flywheel that enables self-reinforcing feedback loops and unlocks faster iteration speeds and higher-quality models.