The specific visuals and details for this project are confidential.

Tools behind the scenes of Google Search

UX for a critical internal platform for Google’s Knowledge Graph

Google | Interaction Designer | 2022 - 2024

What is it

The accuracy of Google’s Knowledge Graph and Knowledge Panels is essential for public trust.

While automated tools serve as the first safeguard, a sophisticated human-in-the-loop approach is critical for validating high-impact information during events like elections, health crises, and breaking news..


I led the end-to-end UX for “Project Veritas” (internal codename), the core platform at the heart of this process.

The challenge was to overhaul the platform into a scalable, precise, and high-performance tool that supports a global curation team, directly improving the efficiency of the information curation workflow.

What I did

As the lone UX designer, I led the full design strategy and rollout. My core task was the evolution of the platform from a manual platform into an efficient, intelligent tool for curation experts.

LLM/AI integration

Conceived and prototyped designs using LLM and AI capabilities.

End to end workflows

Shaped the end-to-end workflow, from creation to assignment.

Design language

Established a unified system of previously missing colors, typography, and interactions, adding consistency and clarity to the curator workflow. 

Universal taskbar

This core element features AI-powered alerts.

Data visualization

Dashboards that enabled managers to tackle quality and schedule risks proactively. 

What comes next is a representative summary of the project, omitting sensitive details to protect confidentiality.

A story of impact: How my work ensured accuracy

To illustrate the impact of my work, let’s picture a hypothetical scenario in San Francisco.

After: New intelligent workflow

  1. A user spots an incorrect population stat in the San Francisco Knowledge Panel.

  2. They flag the error with a “Feedback” submission.

The feedback is auto-analyzed and delivered to the right curator in seconds. The modular taskbar I built delivers AI-powered suggestions, enabling fast verifications and approvals in minutes, not days.

This feedback landed in a shared queue, requiring a manager to manually locate the right curator and assign the task, stretching the fix over several days. 

The knowledge graph is updated only after a manager manually routes the task is assigned to an expert.

Before: Slow, manual correction

  1. A user spots an outdated population stat in San Francisco’s Knowledge Panel.

  2. They report the error using the “Feedback” option.

Outcome & Impact

Drove a substantial increase in task completion.

My taskbar component slashed integration for developers from weeks to just hours.