As the era of large models advances, the landscape of machine learning operations (MLOps) is rapidly evolving, presenting new challenges and opportunities. This transformation encompasses navigating unstructured data, fine-tuning and evaluating complex generative models, as well as deploying and managing inferences on GPU machines. Addressing these challenges requires innovative, scalable solutions that can adapt to the dynamic nature of large models. In this talk, we will delve into our strategies for tackling these emerging challenges, offering insights into our journey and envisioning the future of machine learning and AI in the context of large-scale model operations.