Jason Wu

I’m a Research Scientist in the Human-Centered Machine Intelligence group at Apple. I previously earned my PhD from the Carnegie Mellon Human-Computer Interaction Institute, advised by Professor Jeffrey Bigham. My PhD research was supported by the NSF Graduate Research Fellowship.
Before I joined CMU, I graduated from Georgia Institute of Technology in 2018 with a B.S. in Computer Science. There, I worked at the GT Ubicomp Lab and the Contextual Computing Group.
My research applies computational modeling to understand and improve how humans interact with computers.
featured projects
How can we encode expertise from HCI experts like designers into reward models and use them to improve AI?

UIClip
UIClip is a data-driven model that scores the design quality of user interfaces.

UICoder
Finetuning Large Language Models (LLMs) to generate higher quality UIs through automated compiler and VLM design feedback.

Reinforcement Learning from Designer Feedback
Developing workflows and reward models that encode professional designers' expertise.
news
Sep 26, 2025 | I am attending UIST 2025 in Busan, Korea! Please check out the workshop on longitudinal evaluations of AI systems that I helped organize and SQUIRE. I’d love to talk to any prospective students and collaborators! |
---|---|
Sep 22, 2025 | I am joining Purdue CS as an assistant professor starting Spring 2026. I am looking to work with highly motivated students. Boiler Up! |
Sep 30, 2024 | Thrilled to share that I’ve been promoted from “senior intern” to Research Scientist at Apple! I’ve loved working with the Human-Centered Machine Intelligence group on HCI, UI modeling, and accessibility during my PhD. Excited for what’s next! |
Jun 27, 2024 | Back to Pittsburgh where I passed my PhD dissertation defense! My dissertation, Computational Understanding of User Interfaces would not have been possible without all the amazing support I’ve gotten throughout my PhD. |
Jun 16, 2024 | First time attending NAACL 2024 in Mexico City, Mexico. Amazing city and perfect place to present our work on finetuning LLMs to produce higher-quality UI designs, UICoder. |