Adobe Research
Project Test Kitchen is my Adobe Research internship project, presented at Adobe Summit Sneaks 2026. It uses a novel axis-centered approach to dynamically adjust the design space as designers move from broad exploration to focused refinement. Starting from prompts, reference images, and sketches, designers work with an AI agent to compare and evolve ideas along meaningful design dimensions. The research prototype is called Surprise2Refine.
- Accepted to ACM UIST 2026.
- Presented at Adobe Summit Sneaks 2026, selected as one of seven projects from more than 500 submissions. I was the event's only intern presenter. Read the Adobe Research feature
- Technology transfer into Adobe products is currently in progress.
- Covered by multiple press and news outlets. Explore Project Test Kitchen press coverage.
- A moodboard captures visual references, sketches, and notes to guide generation.
- Two editable conceptual axes organize design variations in a grid.
- Zooming generates closer variations of a chosen design.
- Anchoring blends preferred designs into new alternatives.
- Decomposition extracts reusable visual elements and color palettes for further refinement.
Exploration and refinement in Surprise2Refine. Figure 1, You et al., 2026.
From a moodboard to a finished comedy poster: exploration and refinement in Surprise2Refine.
Comedian Iliza Shlesinger trying Project Test Kitchen live at Adobe Summit Sneaks 2026.
Presenting Project Test Kitchen at Adobe Summit Sneaks 2026.
Watch Project Test Kitchen in action! I presented it live at Adobe Summit Sneaks 2026, where comedian Iliza Shlesinger tried it out on stage.
- Human-AI Co-Creation
- Generative AI
- Creative Scaffolding
- Interaction Design
- AI Agents
Yuzhe You, Gromit Yeuk-Yin Chan, Shunan Guo, Anlan Zhang, Eunyee Koh, Jian Zhao, Tongyu Zhou