Hybrid modeling – the combination of first-principle models and machine learning – offers the potential to increase model accuracy while reducing modeling effort. Although approaches for creating hybrid models from system simulation models exist, the unique characteristics of Modelica-based, object-oriented models – such as modularity and reusability – can, as of today, not be utilized. In this contribution, we explore approaches for bridging this gap to enable the use of hybrid models with Modelica. Key challenges of architecture definition, training environment and reintegration of the trained machine learning parts into a Modelica model are addressed. To illustrate our approach, we present a case study involving a SCARA robot. This example demonstrates a partially integrated workflow for hybrid modeling, intended to serve as a foundation and motivation for further research.