Worked on enhancing model lifecycle reliability and artifact traceability in the mosaicml/llm-foundry repository by implementing MLFlow-based model registration and a multiprocessing saving flow. Leveraged Python and MLOps practices to refactor the registration process, utilizing MLFlow’s log_model API for centralized artifact tracking and improved licensing compliance. Introduced a helper function to handle model saving and registration in a separate process, which reduced resource contention and addressed issues with duplicate tokenizer files and license logging. Updated the test suite to align with the new API and saving flow, resulting in improved continuous integration reliability and more robust test coverage for model deployment workflows.