Being in a place where artificial intelligence  (AI) research is moving fast and where you are working on problems that really matter is just one of the things that Connor Dunlop likes about being a Ph.D. student in computer science at the Sanghani Center. 

“I also really value the opportunities for collaboration and being surrounded by peers and faculty who are both highly skilled and genuinely interested in the work," he said. "And I appreciate having the flexibility to pursue research questions I find important and interesting.” 

Advised by Pinar Yanardag at the Sanghani Center, Dunlop's  research focuses on controllable generative AI, particularly dealing with image and video generation, agentic systems, and preference optimization methods. 

For example, he has worked on a multi-agent framework where specialized agents (e.g. a “creative director” or “art critic”) collaborate to iteratively plan, generate, and refine creative content with minimal user effort. He has also explored personalized image editing, where the system learns an individual user’s aesthetic preferences and adapts edits to match their style consistently.

“As generative models have improved, I have been drawn to the gap between what these systems can produce and how much control users actually have over the output,” Dunlop said. “Generative models can produce impressive images and videos, but it’s often hard to steer them toward what you actually want. That led me to focus on controllability, agentic approaches, and preference modeling, ways to make generative systems more reliable, interpretable, and aligned with user intent rather than just producing plausible results.”

Virginia Tech’s status as a leading research institution with a strong culture of publishing and research impact was a major draw for Dunlop when looking for a graduate program. He was also  personally connected to Blacksburg, having lived there when his dad was pursing his Ph.D. at Virginia Tech. 

“I already knew I loved the area and community,” he said. “And the Sanghani Center specifically felt like a great fit because there are many faculty doing interesting AI and data science work that I am excited to collaborate with.

Dunlop’s collaborative published work includes:

·      “MotionFlow: Attention-Driven Motion Transfer in Video Diffusion Models” at Association for the Advancement of Artificial Intelligence (AAAI) 2026

·      “Personalized Image Editing in Text-to-Image Diffusion Models via Collaborative Direct Preference Optimization” at NeurIPS 2025

·      “CREA: A Collaborative Multi-Agent Framework for Creative Image Editing and Generation” at NeurIPS 2025

Dunlop earned a bachelor of science degree, with a minor in statistics, from the University of Maryland, College Park. Projected to graduate in 2028, he would like to be in an industry position as an AI researcher.