A data engineer needs to build and train a machine learning model using TensorFlow and deploy it for real-time inference with minimal infrastructure management. The engineer wants to take advantage of a fully managed environment to automatically scale resources during model training. Which AWS service should the engineer use?