The MLOPs team transitions machine learning model prototypes to production on NASA High End Computing resources. Developing machine learning model prototypes is an iterative, experimental process involving the manual exploration and pre-processing of data, exploration of hyperparameter space during model training, and verification and validation activities prior to model deployment. This iterative process does not end, however, after model deployment. MLOps is the practice of deploying machine learning models reliably by implementing processes for regular (often continuous) integration of improvements to models and data, monitoring for model drift, and retraining models to ensure that their performance remains within acceptable limits.
