Duygu Ataman

Duygu Ataman

Assistant Professor at Middle East Technical University

Senior Researcher in Accessible & Interpretable AI

Duygu Ataman

About

I am an assistant professor and principal investigator at the Graduate School of Informatics, Middle East Technical University, where I conduct my research on developing efficient, safe and accessible AI systems. Previously, I was an assistant professor and faculty fellow at New York University, Courant Institute of Mathematical Sciences and a post-doctoral researcher at the University of Zürich, Institute for Computational Linguistics. I completed my Ph.D. on neural sequence modeling for applications in machine translation at the University of Trento, and I was also a visiting post-graduate research student at the School of Informatics, University of Edinburgh. In my master's degree in electrical engineering at the University of Leuven I developed analysis techniques with EEG for studying speech perception in the human brain. I hold a bachelor's degree in electrical and electronics engineering from Middle East Technical University.

Research Topics

I am interested in a wide range of topics around self-supervised learning, representation theory, computational modeling of language, reasoning and the theory of mind. I have also contributed significantly to the assessment and development of generalizable and efficient language modeling architectures in order to make language technology accessible across the world, including economically disadvantaged locations where under-represented or under-resourced languages are spoken. In 2021 I have founded the ACL Special Interest Group on Turkic Languages, and since 2020 I co-organize the Multilingual Representation Learning Workshop at EMNLP. I am grateful that my research has been supported by research awards from Microsoft and Google. I am also a recipient of the 2025 TÜBİTAK 2232-A International Fellowship for Outstanding Researchers.

  • Efficient Models

    Model compression and optimization for edge AI and resource constrained settings.

  • Interpretable AI

    Methods for ensuring interpretability and trustworthiness in language models.

  • Multilinguality

    Making technology accessible for under-resourced languages.

  • Multimodality

    Representation learning in the visual domain for applications in image understanding and vision-language models.

Research Group

At the Graduate School of Informatics, I lead a research group of 10 members. My group members are:

I am also affiliated with the METU-BILTIR Center (METU-Computer Aided Design, Manufacturing and Robotic Research and Application Center) where we collaborate with the Department of Electrical and Electronics Engineering to develop the next-generation of AI-integrated industrial robotics.

Teaching

Contact

Email: dataman@metu.edu.tr

Links: GitHub · Google Scholar

Open Positions

Fully funded PhD and MSc positions available — please reach out with CV and a brief research statement.