About me
I'm a postdoc in Noah's Ark at UW NLP, working with Noah Smith. My research focuses on computational auditory scene analysis and music language modeling, heavily inspired by Gestalt psychology and the role of inductive bias in multimodal artificial intelligence. Most recently, I completed a long-term project on building human-readable music notation with audio grounding, designed to provide inductive bias to machine learning algorithms — you can see how it enables a speech recognition model to generate time-aligned sheet music.
Research interests
-
Automatic Music Transcription
Unraveling musical notes from real-life classical music recordings.
-
Music Performance Analysis
Analyzing technical and expressive nuances in musical performances.
-
Auditory Stream Segregation
Untangling complex audio mixtures into individual source parameters.
-
Multimodal Content Retrieval
Extracting information from text, video, and audio for complex tasks.
Personal Interests
Experience & Education
-
Postdoctoral Researcher in Computer Science
Paul G. Allen School — University of Washington 2025 — presentMusic transcription, music language modeling, and computational auditory scene analysis.
-
PhD in Information & Communication Technologies
Music Technology Group — Universitat Pompeu Fabra 2020 — 2024Self- and weakly-supervised learning for various music information retrieval tasks such as music transcription and source separation.
-
M.S. in Electrical & Electronics Engineering
Bogazici University 2017 — 2020Weakly-supervised learning for sign language recognition and keyword search.
-
B.S. in Electrical & Electronics Engineering
Bogazici University 2012 — 2017Specialization in Digital Signal Processing
Languages
-
Turkish
Native -
English
Profficient -
Spanish
Upper Intermediate -
Chinese
Intermediate -
Python
Profficient