Google Research & DeepMind / Internal tools

Designing research tools as AI changed the scope.

For two years, I was the sole UX designer across PubTools and Google Research Hub, two shipped internal applications. I owned the experience from problem framing through design delivery. As AI-related requirements changed, my work expanded from individual workflows into a broader product plan.

Role: Sole UX designer · Contract
Duration: 2 years
Products: PubTools & Google Research Hub

About the recreated visuals: The original screens are confidential. These illustrations were created for this portfolio, with the visual structure informed by a reference from my work. Layouts, labels, workflow states, and data are illustrative. The AI interaction shown is a newly created concept illustrating the exploration described below.

Recreated research-work detail view with a blue application header, structured fields, linked milestones, and status information — Recreated portfolio illustration inspired by the interface’s visual structure. Fields, labels, and content are illustrative.

01 / The challenge

One project grew into several connected workstreams.

The engagement began with one project and expanded across PubTools and Google Research Hub. The work brought together research tracking, request management, dashboards, analysis tools, and complex rules that shaped what users could do next.

As AI-related requirements evolved, the scope kept changing. Addressing each request individually was no longer enough to give the work a coherent direction. The challenge grew from designing individual experiences to understanding how the workstreams should fit together.

02 / My role

Establish the approach, then carry it through.

I was the sole UX designer supporting both applications. Working closely with a project manager, I helped turn ambiguous requirements into UX priorities and a design approach, then developed flows, interaction patterns, and high-fidelity designs.

I partnered with Product, Engineering, and Research to align requirements and translate complex business rules into usable experiences. This internal-tool setting gave me broad responsibility for defining the work as well as delivering it.

Hands-on design

Flows, dashboards, analysis tools, policy-driven resolution experiences, and reusable components and states.

Growing planning responsibility

Clarifying priorities, connecting related workstreams, and developing a broader project plan as the engagement expanded.

03 / Workflow design

Design the states between screens.

My work included high-fidelity dashboards, analysis tools, and policy-driven resolution flows. These experiences connected information users needed to review with the actions available to them.

I also extended the design system with reusable components and states for research tracking and request management. This gave related workflows a shared foundation as the application continued to evolve.

Recreated design board with connected draft, review, confirmation, and update screens — Recreated design-board illustration inspired by my reference image. The screen sequence, states, and content illustrate connected workflow design.

04 / AI exploration

Explore how Gemini could help users decide what needs attention.

Alongside the core workflows, I explored Gemini-assisted recommendations and summaries to help users prioritize actions. This introduced a new layer to the design problem: how an AI-assisted experience could help people make sense of their work and move toward a useful next step.

The recreated concept below illustrates that idea. A short summary highlights a sample item needing attention and connects a suggested action to the underlying request context. This interaction was created for the portfolio to make the exploration tangible.

Recreated AI-assisted concept with an illustrative summary, suggested next action, and a link to request context — AI concept recreated for this portfolio. Illustrative content and behavior show Gemini-assisted summary and recommendation themes; this is not an original screen or a claim that this exact interaction shipped.

What the concept makes visible

A concise summary: bring relevant context into view before asking the user to act.
A connected recommendation: show the work item behind the suggested next step.
A reviewable action: let the user open the context and make the decision.

05 / Product direction

Step back when the scope needs a bigger picture.

The changing requirements made the need for a broader plan increasingly clear. I began taking on project-planning and coordination responsibilities alongside design, connecting the growing workstreams rather than treating every request as a separate addition.

Before my engagement ended, I was developing an overarching project plan to bring the work together. The aim was to give the expanding scope a coherent direction as AI continued to shape the requirements.

Shipped products, with the next phase taking shape

PubTools and Google Research Hub were shipped internal applications. Across the two-year engagement, my contribution spanned end-to-end UX ownership, complex workflow design, reusable patterns, and AI exploration. By the end, that responsibility had grown to include planning the bigger picture for what came next.