My research focuses on efficient vision-language models and screen understanding for computer-use agents.
I am a PhD student in Computer Science at ETH Zurich in the Computer Vision and Geometry Group, advised by Prof. Marc Pollefeys. In parallel, I am a Pre-Doctoral Researcher at IBM Research Zurich on the Docling team, with Dr. Peter W. J. Staar. Before that, I received my M.Sc. in Computer Science from EPFL and my B.Sc. in Computer Engineering from METU.
Released DeskForge, a controllable desktop environment for computer-use agents, with the DeskForge-1M dataset and fine-tuned models.
Invited talk on complete screen parsing supervision at the Microsoft Research Zurich Spatial AI Lab.
Attended the Deep Learning for Science Summer School at Berkeley Lab.
End-to-end key-value extraction from document images accepted at ICDAR 2026.
ScreenParse accepted at ICML 2026.
Started my PhD at ETH Zurich (CVG) and joined IBM Research Zurich as a Pre-Doctoral Researcher.
SmolDocling accepted at ICCV 2025.
Completed my M.Sc. in Computer Science at EPFL.
arXiv 2026
A controllable desktop environment that composes and explores real applications, and DeskForge-1M: 1.2M densely annotated desktop screenshots that improve GUI grounding and long-horizon computer use.
ICML 2026
A large-scale dataset and training framework for complete screen parsing: every UI element of a web screenshot with its box and semantic label.
ICDAR 2026
One model that identifies, locates and links key-value pairs in visually rich documents, without an external OCR step.
ICCV 2025
A 256M-parameter vision-language model that converts whole documents end to end into DocTags markup, competing with models up to 27× larger.
Pre-Doctoral Researcher · IBM Research Zurich
Docling team · with Dr. Peter W. J. Staar
Master's Thesis Student · IBM Research Zurich
Key-value extraction with vision-language models; contributed to SmolDocling and Docling
Research Assistant · EPFL, Mathis Group
Information extraction with large language models · with Alexander Mathis
Machine Learning Engineer · Panoculum Inc.
Deep learning pipeline for an AI-assisted family story-sharing app
Undergraduate Researcher · University of Cambridge
CNN-based blackgrass detection from multispectral imaging · with Cengiz Öztireli
Undergraduate Researcher · University of Illinois Urbana-Champaign
Automated bug localization · with Reyhaneh Jabbarvand
Teaching Assistant · Informatik II, ETH Zurich
Teaching Assistant · Machine Learning (CS-433), EPFL
Ph.D. in Computer Science · ETH Zurich
Computer Vision and Geometry Group · advised by Prof. Marc Pollefeys
M.Sc. in Computer Science · EPFL
B.Sc. in Computer Engineering · Middle East Technical University
