Neural network based machine learning models are commonly used to solve challenging natural language processing (NLP) problems. In the first part of this lecture, we will take a look at large pre-trained language models and how to prompt them to generate specific outputs. We will show a practical application of this in the legal domain, using data form the European Court of Human Rights and the Swiss Federal Supreme Court. In the second part of the presentation we will focus on how to build production software components with machine learning models. We will cover various aspects of MLOps and also challenges around large neural network models.