Chemical and biological processes are inherently characterized by strong and unknown nonlinearities and measurement noise. In addition, the variables from these processes are irregularly sampled and exhibit multiple timescales. A variety of system identification algorithms have been in use in the industry for approximating such processes. However, most of these algorithms do not provide satisfactory performance when accounting for these characteristic features. This chapter will briefly introduce deep learning and its variants as a modeling tool. Deep learning offers an attractive alternative to modeling large-scale processes with sufficiently large data sets. In particular, it can account for the process characteristics mentioned earlier. We will also introduce variants of deep learning, such as recurrent neural networks and variational autoencoders, as tools for large-scale process modeling.