On June 24 the first workshop on probabilistic programming in Hannover took place. Nearly 20 participants from IT companies and the Leibniz University in Hannover came together in the rooms of Data Assessment Solutions (DAS) to discuss this burgeoning topic in machine learning. At the beginning, Joachim Giesen from DAS Research gave an introduction into the ideas behind probabilistic programming, which he distinguished in particular from deep learning. Probabilistic programming languages are flexible modelling tools that free the user from having to worry about inference. Roughly speaking, a probabilistic program is a specification of how to go from parameters to observable data. By means of inference one then infers the values of the parameters from actually observed data. In contrast, in deep learning one specifies directly how to get from observed data to parameters. The direct specification is often much less intuitive than the specification of a data generating process in the form of a probabilistic program. Following the talk by Joachim Giesen, Daniel Borcherding from DAS gave a small practical introduction to the probabilistic programming framework Infer.NET from Microsoft. Using the example of a Topic Model, which is used at DAS for the analysis of employee skills data, he showed what is currently possible and where there are still problems.