I am an Applied Scientist at Amazon Search in Berlin. Previously, I was a NLP researcher and senior staff member at UKP Lab, TU Darmstadt. I am interested in applying NLP research to real-world scenarios, e.g., investigating how to obtain efficient models and studying how to transfer models across different tasks, domains, and languages.
To this end, in AdapterHub, we introduce a novel framework for sharing and integrating parameter-efficient adapter modules in transformers. With AdapterDrop, we establish and improve their computational efficiency. In MultiCQA, we study the zero-shot transfer capabilities of text matching models on a massive scale with 140 source domains to question similarity and answer selection tasks.
In the past, I managed the projects QA-EduInf: Question Answering for Educational Information and SISW: Intelligent Search in the Social Web at UKP Lab, supervised four internships, six BSc/MSc theses, and taught the data analysis project class (in 2016 and 2019) as well as the text analytics research seminar (in 2017). Between 2008 and 2017, I worked as a freelance software engineer for direct clients and medium to large advertising firms such as SYZYGY, Ogilvy, and ETECTURE. In 2011 I founded the company ARULABS UG to carry out larger projects directly.