Machine learning models are difficult to operationalize at scale due to infrastructure challenges like supporting different frameworks and languages, managing models through versioning and reproducibility, and deploying models at large scale. Most organizations struggle with successfully moving projects from proof-of-concept to production as lack of process, incentives, skills, champions, and appropriate technology impede operationalization. Adopting practices like integrating engineering and data science teams, defining clear production criteria, and choosing infrastructure-agnostic platforms can help organizations realize value from machine learning by addressing these barriers to operationalization.