About me

I am a Postdoctoral Researcher at the Applied Machine Learning Lab at the University of Bonn. My work sits at the intersection of NLP and Machine Learning, with a focus on extracting structured knowledge from complex documents and developing autonomous agents for strategic environments.

I hold a PhD in Computer Science from the University of Bonn, specializing in Information Extraction and Natural Language Processing. My doctoral research has extensively explored the application and improvement of transformer-based models to various domains, addressing challenges such as named entity recognition, contradiction detection, automated compliance verification, and sensitive data anonymization.

While I am not doing research, I am responsible for our website, server, and general IT related things.

Before my postdoctoral position I was a Lead Data Scientist at Fraunhofer IAIS, and before that a Data Scientist at EY.

Research interests

• Agentic Systems & Game Theory: Developing autonomous agents capable of reasoning, tool use, and strategic interaction in complex environments (e.g., financial markets, games, or multi-agent simulations).

• Financial NLP: Automating the analysis of financial reports, including KPI extraction, contradiction detection, and regulatory compliance verification.

• Information Extraction: Designing architectures for named entity recognition (NER) and relation extraction (RE) in specialized domains.

• Large Language Models: Investigating the reasoning capabilities, faithfulness, and evaluation of LLMs in high-stakes applications.

Highlighted publications

Google Scholar

• Resource-Efficient Anonymization of Textual Data via Knowledge Distillation from Large Language Models
  Tobias Deußer, Max Hahnbück, Tobias Uelwer, Cong Zhao, Christian Bauckhage, Rafet Sifa
  Proc. COLING, 2025
  [Link]

• Informed Named Entity Recognition Decoding for Generative Language Models
  Tobias Deußer, Lars Hillebrand, Christian Bauckhage, Rafet Sifa
  Proc. BigData, 2024
  [DOI]

• Fusing Speech and Language Models for Dementia Detection
  Tobias Deußer, Abdul Mohsin Siddiqi, Lorenz Sparrenberg, Tobias Adams, Christian Bauckhage, Rafet Sifa
  Proc. BigData, 2024
  [DOI]

• Uncovering Inconsistencies and Contradictions in Financial Reports using Large Language Models
  Tobias Deußer, David Leonhard, Lars Hillebrand, Armin Berger, Mohamed Khaled, Sarah Heiden, Tim Dilmaghani, Bernd Kliem, Rüdiger Loitz, Christian Bauckhage, Rafet Sifa
  Proc. BigData, 2023
  [DOI]

• Controlled Randomness Improves the Performance of Transformer Models
  Tobias Deußer, Cong Zhao, Wolfgang Krämer, David Leonhard, Christian Bauckhage, Rafet Sifa
  Proc. ICMLA, 2023
  [DOI]

• Contradiction detection in financial reports
  Tobias Deußer, Maren Pielka, Lisa Pucknat, Basil Jacob, Tim Dilmaghani, Mahdis Nourimand, Bernd Kliem, Rüdiger Loitz, Christian Bauckhage, Rafet Sifa
  Proc. NLDL, 2023
  [DOI]

• KPI-EDGAR: A Novel Dataset and Accompanying Metric for Relation Extraction from Financial Documents
  Tobias Deußer, Syed Musharraf Ali, Lars Hillebrand, Desiana Nurchalifah, Basil Jacob, Christian Bauckhage, Rafet Sifa
  Proc. ICMLA, 2022
  [DOI]

Contact Details

If you want to get in contact, write an Email to tdeusser[at]uni-bonn.de. You can also swing by my office in Room 2.109, Friedrich-Hirzebruch-Allee 8, 53115 Bonn.