This project extended our Smart Food Delivery project last year with on cloud machine learning function for delivery time prediction and automatic cleaning alert services. Instead of hardware construction, we focused on data collection and analysis, as well as usability and reliability of our products this semester. With Parse service closing, we migrated our mobile app database to AWS DynamoDB, and deployed AWS machine learning triggered by Lambda to learn the predicted delivery time for each order in order to save energy and maintain taste of food. The result will then be sent to costumer by AWS SNS as a push notification. In order to remain the delivery box clean, we did 100 experiments with spill and no spill in box, and compare their difference in terns of the change in humidity. The results was used for classification by AWS machine learning, and the system is now able to judge if there is leakage or spill in the box, and can automatically call cleaning services to ensure the box is in clean condition.