Peyman Gholami

Peyman Gholami

I'm a final-year PhD candidate in Computer Science at the University of British Columbia, advised by Leonid Sigal and Robert Xiao. My research spans Artificial Intelligence, Human-Computer Interaction, and the full spectrum between these two. What I am really interested in is applying AI and multi-modal learning techniques to interactive and creative applications.

My PhD work focuses on Selective Intervention for Efficient and Controllable Visual Foundation Models, which addresses the shortcomings that limit these models in practice by finding where targeted changes make the biggest difference. Along the way, I've had the chance to work closely with leading researchers at Google DeepMind, Meta and Microsoft Research. Before that I was a research scientist at Mirametrix, building face and gaze tracking systems, and did my M.Sc. at the University of Waterloo.

I'm on the job market for research positions starting in early 2027.

Selected publications

DISCOVER timeline: REPR probes, guided selection, growing coverage and frozen reweighting over GRPO training
NeurIPS, 2026

Decides which prompts are worth labeling while a vision-language model trains, focusing on the ones it still answers inconsistently. With only 10% of the data labeled, it outperforms GRPO trained on the full pool for the same number of steps.

ICCV, 2025  (Best Demo Award)

Brush-based image editing with diffusion models, where every edit lives on its own layer and can be hidden, adjusted or removed later. No training required, and edits appear in real time.

All projects  /  Full list on Google Scholar

Experience

Education

Honors

Teaching and service

Teaching assistant at UBC for CPSC 340 Machine Learning (five terms, lead TA 2022), CPSC 540 Advanced Machine Learning (four terms), CPSC 330 Applied Machine Learning (lead TA 2023), CPSC 100 Computational Thinking (lead TA 2022), CPSC 425 Computer Vision and CPSC 221 Algorithms and Data Structures.

Reviewer for NeurIPS, ICCV, ECCV, WACV and UIST.