Coding is an important process in qualitative research. However, qualitative coding is highly time-consuming even for small datasets. To accelerate this process, qualitative coding systems increasingly utilize machine learning (ML) to automatically recommend codes. Existing literature on ML-assisted coding reveals two major issues: (1) ML model training is not well integrated into the qualitative coding process, and (2) code recommendations need to be presented in a trustworthy way. We believe that the recently introduced concept of interactive machine learning (IML) is able to address these issues. We present an ongoing design science research project to design an IML system for qualitative coding. First, we discover several issues that hinder the success of current ML-based coding systems. Drawing on results from multiple fields, we derive meta-requirements, propose design principles and an initial prototype. Thereby, we contribute with design knowledge for the intelligent augmentation of qualitative coding systems to increase coding productivity.