What is your understanding of the term model drift, and how would you address it for a deployed ML model at Cisco?
