Ph.D. Candidate in Computer Science, Virginia Tech
I am a Ph.D. candidate at Virginia Tech, advised by Prof. Pinar Yanardag. I study the internal representations of generative models, how they encode meaning, structure, and style, to make image and video generation more interpretable and controllable.
My recent work spans controllable generation, diffusion model distillation, and mechanistic interpretability for generative models. I want to understand why these systems behave the way they do, not just improve their outputs.
My papers have appeared at CVPR, ICCV, ICML, and AAAI. Before my PhD, I spent several years building production machine learning systems. I have also interned at Waymo, Amazon, and Adobe during my PhD.
Selected Highlights
- Publications at CVPR 2026, AAAI 2026, ICCV 2025 (Highlight), ICML 2025 (Oral), and CVPR 2024.
- Research internships at Waymo (2026), Amazon (2025), and Adobe (2024).
- Industry deployment experience with diffusion-based services serving over 5 million daily requests.
- Co-organizer of the ICCV 2025 and CVPR 2026 Personalization in Generative AI (P13N) workshops.
Current and Recent Roles
- Waymo - Research Intern (Incoming, Summer 2026)
- Virginia Tech - Research Assistant and Lab Lead (Aug 2023 - Present)
- Amazon - Applied Scientist Intern (May 2025 - Aug 2025)
- Adobe - Research Intern, Video Generative AI (May 2024 - Aug 2024)
- Lyrebird Studio - Machine Learning Engineer (Nov 2022 - Aug 2023)
- Vispera - Computer Vision Research Engineer (Oct 2019 - Nov 2022)
Blog
Generating Pixels One by One
Your First Autoregressive Image Generation Model
Recent News
Feb 2026
Passed the Ph.D. preliminary exam and advanced to candidacy. Dissertation: Representation Steering as a Control Surface for Generative Foundation Models.
Feb 2026
Infinity-RoPE was accepted to CVPR 2026.
Dec 2025
The CVPR 2026 Personalization in Generative AI (P13N) workshop proposal was accepted in December. This is the second workshop edition.
Nov 2025
Released the Infinity-RoPE preprint on arXiv.
Selected Publications
MotionFlow: Attention-Driven Motion Transfer in Video Diffusion Models
AAAI 2026 (Main Technical Track) · 2026
CLoRA: A Contrastive Approach to Compose Multiple LoRA Models
ICCV 2025 (Highlight) · 2025
ConceptAttention: Diffusion Transformers Learn Highly Interpretable Features
ICML 2025 (Oral) · 2025