How would you architect the deployment pipeline for a new sentiment analysis model, ensuring zero downtime and continuous data validation?
