felixpeters Goto Github PK

I'm a data scientist at Heraeus, working on incorporating AI into the manufacturing workflow. Formerly, I was an AI researcher in the Software & Digital Business Group at Technical University of Darmstadt.

My main research interests are human-AI interaction (especially, explainable AI) and applications of AI in the healthcare sector, in particular medicine. For a list of my publications, see my Google Scholar profile. Among my research highlights are:

• Developing an ML model to segment and classify lung nodules from chest CT scans (see repo). We used this model in a web application to test whether explainable AI features (e.g., heatmaps, example-based explanations) increase transparency for the physician, while not increasing his perceived cognitive effort (see repo).
• Implementing an agent-based simulation model in order to examine how increasing human-AI interaction affects organizational learning (see repo). The accompanying article was published in the A+ journal MIS Quarterly.

Ever since completing my first ML course during my Master's study back in 2015, I have been interested in all things ML engineering and data science. Lately, I've dived a lot into MLOps, i.e., the automation of data and ML pipelines. Some of my personal work highlights include:

• Implementing and deploying deep learning models for visual inspection and key information extraction.
• Delivering in-person and virtual ML workshops at a large financial institution and a employers' association from the manufacturing industry (see repo).
• Completing two data science projects with the sales department of a large public transport company.
• Completing multiple side projects, e.g., visual inspections of capsules (see repo) and predicting tweet engagement (see repo).