Duygu Ataman, Ph.D.
Senior AI Scientist and Engineer
About me
I am a scientist and engineer with over 10 years of experience working in cutting-edge research and development projects across the industry and academia.
My current academic research focuses on achieving structural generalization in artificial intelligence (AI) systems. More specifically, I am interested in building low-cost, efficient and reliable machine learning algorithms that could use the same reasoning principles to solve a wide range of complex problems. In addition to designing novel deep learning architectures and optimization methods, I also actively work on developing evaluation benchmarks and measures that can assess the generalization capability of systems across modalities (e.g. vision to text), languages, or different types of inference tasks.
Education
Ph.D. in Information Engineering and Computer Science
Università degli Studi di Trento, Italy
2019
Designed and developed open-vocabulary generalizable language generation models through self-supervised learning methods applicable across languages, such as statistical tokenization algorithms, compositional character-based language models and Bayesian variational inference methods.
Thesis advisor: Marcello Federico
M.Sc. in Electrical Engineering
Katholieke Universiteit Leuven, Belgium
2015
Graduated with a specialization on Embedded Systems and Multimedia Technologies. My master’s thesis conducted in collaboration with the Department of Neurosciences studied the effect of attention on the perception of speech in the human brain using EEG-based signal reconstruction methods.
Thesis advisor: Marc Moonen
B.Sc. in Electrical and Electronics Engineering
Middle East Technical University, Turkey
2013
Graduated with a minor focus on Computer Science. I was a member of the Renewable Energy Society and a part of the 3rd ranked team in the national solar car races across universities of Turkey in 2011. I was also a member of the Robotics Society and in the organization team of International Robotics Days.
Undergraduate advisor: Aydın Alatan
Experience and Scientific Training
Assistant Professor and Faculty Fellow
Courant Institute of Mathematical Sciences, New York University (2021 - Present)
Since 2021, I am a faculty fellow at the Courant Institute of Mathematical Sciences which allows me to conduct my independent research on developing novel modeling architectures for generalizable and efficient generative language models. I frequently supervise research students that contribute to my ongoing research activities in various directions, most of which recently included design and development of novel multilingual language modeling methods, developing evaluation measures and metrics for assessing the generalization capability of generative models across languages and modalities, and new benchmarks presenting challenging tasks for assessing different inference capabilities. As an assistant professor at the Department of Computer Science, I also taught the major courses Introduction to Computer Science, introducing object-oriented programming with Java, and Foundations of Machine Learning, where I present a comprehensive and practical machine learning course including practical implementations with Pytorch.
Post-doctoral Researcher and Lecturer
Institute for Computational Linguistics, University of Zürich (2019 - 2021)
I was a senior researcher in the MUTAMUR project directed by Prof. Rico Sennrich on extending the applicability of language models in data-scarce languages and tasks using multi-task, multilingual and multi-modal learning methods. During this time, in addition to research activities, I also started the organization of the Workshop on Multilingual Representation Learning (in collocation with the Conference on Empirical Methods in Natural Language Processing) to promote research in this field. I also designed and taught a new course in the undergraduate program on the Creation and Annotation of Linguistic Resources, which provides an introduction to natural language processing with a focus on data collection, annotation and processing methods. At the end of the course students prepared corpus resources in European languages, in addition to Swiss dialects, and contributed to the literature of language resources in Switzerland.
Applied Scientist Intern
Amazon (2019)
Details confidential.
Visiting Post-Graduate Research Student
School of Informatics, University of Edinburgh (2018)
I visited the statistical machine translation group at the University of Edinburgh for five months in 2018 to study under the supervision of Prof. Alexandra Birch, where I contributed to the SUMMA project, a collaboration with BBC and Deutsche Welle for developing efficient real-time machine translation technology for multilingual media monitoring.
Research Fellow
Statistical Machine Translation Research Group, Fondazione Bruno Kessler (2015 - 2018)
Research fellowship awarded for designing and developing efficient multilingual representation learning and neural machine translation models that can be built with minimal data and applicable across languages with various morphosyntactic typologies.
Research Assistant
Department of Neurosciences, Katholieke Universiteit Leuven (2015)
I conducted my master’s thesis research on Electroencephalography (EEG) signal processing and clinical experimentation at the Experimental Otorhinolaryngology research group as part of a collaboration with the Electrical Engineering department for the development of new generation neuro-steered hearing aid devices.
Research Engineer
You Know Watt (2014 - 2015)
During my master’s studies, I was a part-time research engineer at You Know Watt, a well-funded start-up based in Brussels that won many competitions for its innovative smart energy consumption management solution. As part of the core research and engineering team, I implemented the main power signal receiver module of the platform and its embedded software (C) and optimized it for real-time signal processing. I also helped research and development of machine learning algorithms for monitoring and analysis of collected power data.
Project Engineer
Darkblue Communication Systems (2012 - 2013)
I joined Darkblue as a part-time project engineer in my senior year in the scope of a project in collaboration with the Scientific Council of Turkey for the design and development of the first direction finding technology in Turkey, for GSM band signal tracking in search and rescue applications. During this project, I conducted extensive research to design and develop an antenna system and signal receiver circuits, contacted producers and found international partners for hardware development, implemented embedded signal processing modules (C#), designed and implemented the user interface for the device screen and performed field tests with the developed system.
Research Intern
Institute for Information and Communication Technologies, Electronics and Applied Mathematics, Université catholique de Louvain (2012)
In summer of 2012, I was awarded a research grant to visit the research group of Prof. Benoit Macq at ICTEAM, UCLouvain, Belgium, to contribute to their ongoing project on real-time visual human interaction platform with hand gesture based remote control. During this internship, I conducted research and implemented real-time visual feature extraction, signal processing and Kalman filtering methods (C++) for the recognition and tracking of hand gestures.
Awards and Fellowships
2024
Gemini Academic Program Principal Investigator
Research award providing cloud credits for the evaluation of large language models across various generalization tasks.
Microsoft Research Awards: Accelerate Foundation Models Program Principal Investigator
Research award supporting project on increasing the applicability of foundation models across world languages.
2023
New York University, Faculty Fellowship
Fellowship awarded by the Courant Institute of Mathematical Sciences in New York for research on investigating symmetric properties of combinatorial generation problems and integrating these principles into language models to enhance their capability in compositional generalization across languages and tasks.
2021
Spotlight Award, International Conference on Learning Representations
Awarded to research paper ‘A Latent Morphology Model for Open-Vocabulary Neural Machine Translation’
2020
Workshop on Neural Generation and Translation, Travel Scholarship
Scholarship awarded for sponsoring participation in the Workshop on Neural Generation and Translation, at EMNLP to present paper ‘On the Importance of Word Boundaries in Hierarchical Neural Machine Translation’.
2018
Association for Computational Linguistics (ACL), Walker Fund Scholarship
Scholarship awarded for sponsoring participation in the Annual Meeting of the ACL to present conference publication ‘Compositional Representation of Morphologically-rich Input for Neural Machine Translation’.
2018
University of Edinburgh, Research Scholarship
Scholarship awarded to visit the Statistical machine translation group and collaborate on developing a novel neural machine translation model for the translation of low-resourced and morphologically-rich languages.
2018
Università degli Studi di Trento, Doctoral Fellowship
Research fellowship awarded for doctoral research on design and development of novel deep learning algorithms for open-vocabulary language generation.
2015
Erasmus Mobility Grant for Research Internship
Scholarship awarded to be a visiting research student at Université catholique de Louvain in Belgium.
2012