Making new interactions tangible enough to test.

Early generative-AI capabilities created a wide field of possible product behavior. Familiar interface conventions did not automatically explain what Gemini could do, how people might direct it, or what users would need in order to understand and trust the result.

The work required a bridge between technical possibility and product experience: something more specific than a concept deck, but faster and more adaptable than production implementation. It also required new ways to evaluate motion, color, content, and accessibility as the visual language around Gemini developed.

When capabilities and expectations are both changing, a polished demo can create false certainty. The most useful prototypes expose assumptions: how a person begins, what they can create, how the system responds, how a design moves, and whether a new color treatment remains legible.

Each experiment needed enough realism to support meaningful research while remaining flexible enough to change quickly. That meant giving colleagues tools to vary content and presentation without waiting for a UX engineer to rebuild every combination by hand.

Julie partnered with design and motion design to build performant brand animations for the Gemini logo and explore other motion treatments. Her frontend and UX background made it possible to work across visual intent, animation performance, interaction detail, and implementation constraints.

She also made significant contributions to the shared UXE React framework used across the Search organization to build user-research prototypes. Improving that foundation gave UX engineers a more capable common starting point and allowed individual experiments to focus on the question being studied.

Julie built an internal tool that let colleagues change color themes inside Lottie animation files and preview treatments across surfaces. Motion partners could explore a broader range of brand options while keeping the animation asset and its behavior intact.

She also built a color-contrast tool for design colleagues to preview new color families and compare brightness levels against accessibility standards. Instead of discovering contrast problems after a visual direction had hardened, teams could evaluate them while the palette was still being shaped.

Julie built Yo, Google!, an authenticated headless CMS that supported dynamic content creation on top of Search designs. Researchers and collaborators could prepare different content conditions for experiments without embedding every scenario directly in a prototype.

She also built voice-to-content demos using early versions of Gemini. Participants could create AI-generated content through speech, giving teams a tangible way to evaluate user-generated content, authorship flows, and the interaction possibilities of emerging models.