Lukas Galke Poech | Tracing connectionist mechanisms for safer AI systems.

Lukas Galke Poech

Assistant Professor
University of Southern Denmark (SDU)
Department of Mathematics and Computer Science (IMADA)
Data Science Section
Center for Machine Learning

Situated in machine learning and natural language processing, my research aims to make powerful language models more robust, transparent, and controllable. I bring together representation learning and psycholinguistics—linking how language models behave with why they behave that way—to build principled methods for advancing AI Safety.

News

  • November 2025: The MIST project on mechanistic interpretability for safe LLM agents has been funded by the Novo Nordisk Foundation. I am now hiring at PhD and Postdoc level—get in touch if you're interested.
  • November 2025: FlexDeMo has been accepted to AAAI 2026.
  • October 2025: Culture Neurons have been accepted to AACL-IJCNLP 2025 Findings.

Contact: lukas 'at' lpag.de
Design: Adapted from Diane Mounter.
Privacy: No personal data, no cookies.