MLOps is an emerging method to fuse ML with software development by integrating multiple domains. MLOps combines ML, DevOps, and data engineering, aiming to increase the cooperation and working efficiency of AI practitioners, teams, and enterprises by building, deploying, and maintaining ML systems.

Generally, the MLOps pipeline performs operations such as build, deploy, and monitor, which work modularly in sync with each other. The build module is the ML pipeline's core, containing data ingestion, model training, model registration, and model evaluation. The deploy module is for application testing and production release. Finally, the monitor module works in sync with the deploy module, which helps monitor the performance of the model and the deployed applications.

Starwhale starts the MLOps journey with model evaluation because we think model evaluation plays a role in machine learning. Model evaluation quantifies the performance of models on the test dataset. The evaluation metrics help data scientists to understand a machine learning model's performance. Thus, the weakness and strengths of a model are known. By comparing the metrics of models, the best-performed model can be chosen and released to production.