Breaking the Silos Between Data Scientists and DevOps with New MLOps Practices