You are an engineer at a public transportation company responsible for improving the user experience for your customers by providing real-time delay time estimates for multiple transportation routes. The predictions generated by the model must be served directly to users via a mobile app in real time. Considering that different seasons and population increases affect route data relevance, you must retrain the model every month to maintain accuracy. Following Google-recommended best practices for machine learning workflows, how should you design the end-to-end architecture for training and deploying your predictive model?