Drove a +33.7% lift in new editor activation on English Wikipedia mobile web by evaluating a structured Add-a-link feature. The experiment also reduced edit reverts, improving contribution quality at scale.

Analyzed early adoption and quality signals for Reference Check to assess whether the feature was ready for large-scale experimentation.

Evaluated the Reference Check A/B test and translated experiment results.

Uncovered the financial strategies shaping a major election by analyzing tens of thousands of TV ads and political contracts. Translated opaque spending patterns into clear visual stories published nationwide, helping journalists reveal who was influencing the race — and how.

Translated complex demographic data into a clear visual story showing where and how San Francisco’s senior population is changing.

Mapped the organization's digital heartbeat by unifying analytics scattered across tools and teams. Built a dashboard that surfaced the patterns behind traffic surges, conversion shifts, and campaign impact — transforming ad-hoc reporting into a clear, reliable, and daily source of truth.

Tested surfacing edit suggestions directly in read mode. The experiment showed no activation gains, informing iteration rather than rollout.

Automated the reporting backbone that powered fundraising, staffing, and editorial strategy. By pulling data directly from analytics sources into Google Sheets, the system replaced hours of manual searching with instant clarity — enabling faster, more confident decisions.

Reimagined the digital presence of a green-cleaning cooperative network, from branding to user flows to site architecture. The redesign strengthened donor trust, clarified the co-op story, and made it far easier for customers to choose socially responsible services.

Rebuilt critical voter-facing pages to make civic information easier to reach and understand. Paired UX improvements with an optimized AdWords strategy that dramatically expanded reach and lowered acquisition costs, helping a national nonprofit support more voters at key moments.

Transformed a slow mobile site into a near-instant experience, pushing PageSpeed scores to 99/100 and stabilizing animations at 60fps. Clarified the relationship between visual polish and engineering rigor — and proved how much performance shapes perception.

Revealed how subtle page changes shaped browsing, engagement, and purchasing behavior. Brought statistical clarity to a high-stakes product decision by showing not just whether a new design performed differently — but why it mattered.

Illuminated student behavior inside an online learning funnel by analyzing how homepage and feature variations guided learners toward enrollment. Helped the team see which ideas sparked genuine progress and which needed rethinking before launch.

Explored decades of temperature patterns to understand how local climate trends echo — or diverge from — global warming signals. The resulting visual narrative made long-term climate patterns easy to grasp at a glance.

Mapped the rhythm of city movement by analyzing bike-share usage across Chicago, NYC, and D.C. Revealed when, where, and how riders travel — from commuter surges to seasonal shifts — through a guided, interactive Python exploration.

Built an interactive map that layers geolocation with real-time news to give neighborhoods texture and context. The app brings together APIs, filters, and lightweight UI components to help users see a place from multiple angles at once.

Unified machine learning predictions, a Twitter archive, and API metadata to uncover the patterns behind one of the internet's most beloved datasets. The cleaned, enriched dataset set the stage for deeper insights into how audiences interact with social content.

Supported reporting on Oakland's Safe Routes to School by analyzing transportation and demographic patterns. Helped show how infrastructure and geography shape equitable access to education — a story with deep community implications.

Examined San Francisco's opaque school assignment process by analyzing thousands of parent choices and multi-round lottery outcomes. Revealed how family preferences, program structures, and algorithmic constraints interact in a system that shapes educational opportunity.

Mapped 5+ years of NYC collision data to identify crash hotspots and time-based risk patterns across Brooklyn.

Goal: Design and develop a vocabulary learning tool for English language learners.
Audience: English learners, educators, language learning programs.
Approach: User-centered design, iterative testing, educational technology development.
Tools: UX design tools, web development, user testing.