Rama Ramakrishnan continues MIT's 15.773 Hands-On Deep Learning course with a session on training deep neural networks, using a heart disease prediction model as the running example. He covers practical aspects of network design and training, then introduces Keras and TensorFlow as the tools for building and running these models. The session shows how deep learning techniques, often associated with images and text, apply to tabular data of the kind found in medical records and spreadsheets. Ramakrishnan works through the mechanics of setting up a model in code, framing choices around architecture and training against the concrete goal of predicting heart disease from patient data. The lecture runs about 78 minutes and functions as a hands-on continuation of the course's opening material on deep neural network fundamentals.