About me
Hi, I am Ezgi. I am an incoming ML Research Engineer at Apple, where I will be working on problems at the intersection of ML + Video.
I recently completed my PhD in Electrical and Computer Engineering at New York University, advised by Elza Erkip. I hold an (integrated) MEng degree in Electrical Electronics Engineering from Imperial College London. My recent collaborators include Jona Ballé and Aaron B. Wagner in academia, and Zhiqi Chen, Kedar Tatwawadi and Oren Rippel from Apple.
I’m a collaborative researcher who enjoys working across diverse teams. My PhD work focused on bridging theory and practice in data compression by drawing on tools from deep/machine learning, signal processing, and information theory. More recently, I’ve also begun working on perceptual optimization and 3D vision.
I am always happy to connect with researchers at the intersection of ML, neural compression and information theory — feel free to reach out to me at me(at)ezgi(dot)space.
Recent (Selected) Updates
December 2025: I successfully defended my PhD thesis titled "Neural Distributed Data Compression & Communication". Find the slides here!
July 2025: My PhD research was featured on NYU Tandon's website! Many thanks to Mari Rich for the fun interview.
May 2025: Excited to be joining Apple as an ML + Video Research Intern this summer in Cupertino! Relocated to San Francisco/Bay Area for the summer.
March 2025: Honored to be selected as an iREDEFINE 2025 Fellow by the Electrical and Computer Engineering Department Heads Association (ECEDHA)! Here is the poster I presented.
February 2025: Gave an invited talk at the High-Beams Seminar, where I discussed my recent work on learning-based distributed (data) compression. Many thanks to Kaan Aksit for the kind invitation!
January 2025: Our workshop proposal titled "Learn to Compress and Compress to Learn" for 2025 IEEE International Symposium on Information Theory has been accepted! Check out our workshop website here.
December 2024: Honored to be selected as a recipient of the 2024 IEEE Signal Processing Society (SPS) Scholarship!
November 2024: In our recent NeurIPS'24 workshop paper, we discuss a few overarching failure modes of some popular class of neural compressors – that is, their difficulty in learning discontinuous functions.