Thomas
Dethmann

My background bridges the humanities and engineering. I hold degrees in English & Musicology (B.A.), Audio & Video Engineering (B.Eng.), and Audio Communication & Technology (M.Sc.), including a semester abroad at Lawrence Technological University (GPA 4.0, Dean's Honor Roll) in Southfield, MI USA.

Today I work as an ML Research Engineer at Fraunhofer IAIS, developing multi-modal large language models that integrate speech and text for ASR, spoken question answering, and translation (published at ITG Berlin 2025). I also build LLM/ASR-based pipelines — including video summarization for the Lamarr Institute, data augmentation for production ASR systems such as air-traffic control scenarios, prompt engineering for LLM fine-tuning as part of the MWL project, and multilingual translation models for a major German broadcaster.

Previously I interned as a Machine Learning Engineer at Apple, building deep neural networks for automatic drum transcription as part of my award-winning Master's thesis. I also worked at Brainworx Audio on hardware modeling and ML-based instrument classification, and at Fraunhofer HHI on real-time spectrum analysis for drone communications.

I've been making music since age ten and was the drummer of Fil der Protagonist. As a multi-instrumentalist with professional training in recording, mixing, and sound engineering, I also worked as an acoustics consultant — planning room acoustics for offices, schools, and concert halls — and spent three years teaching drum set to students of all ages and skill levels.