Emircan G.

MSc in CS @ ETH Zürich | Ex-Imperial, INSAIT, KAIST | Reconstructing 3D Scenes

Zurich, Zurich, Switzerland

About

Research. 3D computer vision and diffusion models. https://emircangun.github.io

Experience

  • Research Project at Google
    Feb 2025 - Present · 1 yr 5 mos

    • Developing a feed-forward pipeline for reconstructing dynamic 3D human scenes from monocular video under M.-J. Rakotosaona (Google) and D. Barath (ETH).

  • ETH Zürich (Zurich, Switzerland · On-site)
    • Teaching Assistant
      Feb 2025 - Present · 1 yr 5 mos

      • Prepared content for AI and IT in Industry (CAS) and Advanced Machine Learning (MSc) courses. • Prepared material on reinforcement learning, large language models, computer vision, and diffusion models.

    • Research Assistant
      Dec 2024 - Aug 2025 · 9 mos

      Computer Vision and Learning Group under Prof. Siyu Tang and Frano Rajic. • Researched on multi-view 3D point tracking with Gaussian Splatting.

  • Istanbul Technical University (3 yrs 1 mo)
    • Research Assistant
      Jun 2023 - Present · 3 yrs 1 mo

      Vision Lab under the supervision of Prof. Gözde Ünal. • Researched machine unlearning for computer vision tasks, evaluating methods for multi-label scenarios.

    • Research Assistant
      Sep 2023 - Present · 2 yrs 10 mos

      Research group under Dr. Murat Çelik. • Researched applications of machine learning and graph learning in nanomechanics.

  • Visiting Student Researcher at Korea Advanced Institute of Science and Technology
    Jun 2025 - Aug 2025 · 3 mos

    KIXLAB under Prof. Juho Kim. • Worked on text-to-video diffusion model that generates complex scenes through vision–language reasoning.

  • Research Fellow at INSAIT - Institute for Computer Science, Artificial Intelligence and Technology
    Jun 2024 - Sep 2024 · 4 mos

    Collaborated with Prof. Martin Vechev, Dr. Maximilian Baader, and Dimitar Dimitrov from ETH Zurich. • Researched training dynamics in deep learning, focusing on neural collapse, low-rank gradient properties and machine unlearning.