Talk by Katharina Strecker

Title: Research Opportunities for Domain-Expert AutoML Systems

Abstract: AutoML promises to lower the barrier for applying machine learning by automating model selection, hyperparameter tuning, and optimization. However, despite these advantages, many domain experts either struggle to effectively use AutoML tools or choose not to rely on them at all. This talk explores the key limitations of current AutoML systems from a domain expert’s perspective, highlighting critical gaps in usability, interpretability, and integration into existing workflows. In the talk we will discuss possible research opportunities in keeping domain experts actively involved, making AutoML outputs more interpretable for non-experts, and automating preprocessing without sacrificing expert control.