Selected Works
Gemini Enterprise
→ 2024 / 2026
I’ve been designing for the moment when workplace AI stops being a cool demo and becomes a reliable part of how people get work done. Gemini Enterprise is positioned publicly as the “front door” that employees can actually show up to every day, and that admins can roll out without losing governance. That matters because the real barrier is rarely the model. It’s trust, access, and the messy reality of knowledge living across ten different systems. Taken together, the story is simple. A unified entry point. Grounded by connectors. Made real by removing the last-mile friction of “getting the assistant the thing.” The work is design as translation: between employees and admins, between chat and action, and between what people need in the moment and what the enterprise can safely allow.
→ Initial launch
→ Connectors + data stores
→ Agent builder
→ Slack integration
AI search client demos
→ 2023 / 2026
I was designing for two groups at once: Google Cloud sellers and solution engineers who needed a compelling way to tell a product story, and the client UX teams on the other side of the table who needed to understand what “good” could look like for their own platforms. The product was a demo-driven service: a set of north-star narratives, working prototypes, and facilitation sessions that translated Vertex AI Search conversational search v2 into something client teams could see, react to, and build toward. People needed it because conversational and multimodal search can feel abstract until it is grounded in real flows, real content, and real constraints. Sales needed clarity to create momentum and trust. Client UX teams needed a partner who could bridge strategy and implementation so the vision did not die in the handoff. The “wow” was making the future feel tangible. I would bring agentic flows, AI Mode integration, multimodal and media search into prototypes that clients could interact with, then use that shared artifact to align stakeholders, surface risks early, and turn a high-level pitch into an implementation-ready conversation.
→ A2UI
→ Agentic flows
→ AI Mode integration
→ Multimodal search
→ Media search
Merchandising console
→ 2024 / 2025
I’m designing for merchandisers and ecommerce operators who are accountable for revenue and conversion, but don’t want to depend on engineering cycles every time a promotion changes. The product is a Merchandising console inside Vertex AI Search for commerce that lets business users create and manage “controls” (like pinning, boost/bury, synonyms, and facets) and route those changes through an approval flow, with Cloud admins managing access (Creator/Approver roles). Public docs describe this as a way to configure how search/browse/recommendations behave so teams can react faster to trends, stock changes, and business goals—without turning every tuning decision into a code deploy. What’s unique here is the combination of operational governance (approval + roles) with a dedicated UI surface for controls management, which is a trust and speed unlock for real retail teams working under time pressure.
→ Governance & access
→ Control authoring
→ Merchandising controls
→ Discovery tuning
→ Preview & explainability
Vertex AI Search
→ 2023 / 2025
Vertex AI Search is a managed enterprise search platform that lets developers ingest their organization's data, configure how answers are generated, and embed a working AI search experience without training a model from scratch. It's built for enterprise software developers, often skilled engineers with little to no machine learning background, who are under pressure to ship AI-powered search without a six-month build cycle. Behind them are the people actually doing the searching: employees digging through internal knowledge bases, or customers on a company's site. Before this product existed, building something equivalent meant assembling a vector database, fine-tuning an LLM, wiring a retrieval pipeline, handling citation logic, and writing the front end from scratch. Most engineering teams couldn't staff that. The result was either no AI search at all, or something brittle and poorly grounded. Vertex AI Search gave teams a production-ready starting point that already knew how to retrieve, summarize, and cite.
→ Onboarding flow
→ Configuration surface
→ Search experience widget
→ Analytics and evaluation
→ Query & results UI
Recommendations AI
→ 2023 / 2024
Retailers using Google Cloud's Recommendations AI fell into two groups: the engineers who could configure it, and the merchandisers who actually needed it to work for their business. The console was only built for one of them. Between 2023 and 2024, I redesigned three core areas of the product: data quality and catalog health, analytics and reporting, and merchandising controls. The before state was a technically capable product that non-engineers couldn't meaningfully use, with no visibility into why AI performance was lagging and no way to act on it without filing a ticket. The work made the system legible: data readiness became a structured checklist, full-funnel analytics covering page views, cart adds, purchases, and revenue came into the console without needing custom SQL, and merchandising controls were redesigned so a business user could adjust rankings independently. The differentiator wasn't the algorithm. It was making the algorithm trustworthy to the people who weren't running it.
→ Data quality visibility
→ Full-funnel analytics
→ Merchandising controls
→ Performance tier tracking
→ In-console reporting
Document AI
→ 2022 / 2023
I designed the console experience for Document AI Workbench, a Google Cloud tool that helps enterprise teams pull structured data out of documents automatically. The people I designed for (loan officers, compliance reviewers, financial analysts, and the developers supporting them) spent hours manually reading and re-entering information from invoices, contracts, and forms. My work covered the full product: a redesigned Processor Gallery to help users find the right tool faster, labeling flows for training custom models to extract, classify, and split documents, and a new interface for the GenAI Extractor and Summarizer, where users describe what they want, preview results instantly, and correct rather than label from scratch. The core design challenge was that each task required a different way of working, yet all of it had to feel like one product.
→ Workbench GA
→ Processor Gallery revamp
→ Custom Document Classifier
→ Custom Document Splitter
Xero
Workpapers
→ 2020 / 2022
Accountants and bookkeepers at small accounting practices spend compliance season manually copying figures between Xero Workpapers and Excel spreadsheets they have spent years building and trust completely. I designed a two-way sync that connects both tools, so changes made in either place are reflected in the other automatically. The problem was never that accountants were wrong to love Excel. It was that their cloud compliance platform had no way to meet them there. My role covered the full experience: researching how practices actually worked during year-end, defining the sync model, designing the conflict resolution flow, and shipping a global beta. What made it different was the decision not to replace Excel but to treat it as a first-class surface alongside Workpapers, giving practices a single source of truth without asking them to abandon the templates they had built their workflows around.
→ Onboarding flow
→ Microsoft Marketplace materials
→ Excel widget design
→ Workpapers interface
Transamerica
Digital Transformation
→ 2018 / 2020
Transamerica serves millions of Americans across life insurance, annuities, and retirement plans, but its customers had no coherent digital home. Years of growth through acquisition left individual policyholders and retirement plan participants navigating a patchwork of legacy portals, each tied to a different product line rather than to the person holding them. I worked across the digital experience redesign that consolidated those fragmented touchpoints into two purpose-built portals: the Individual Experience for annuity and life insurance customers, and the Participant Experience for workplace retirement plan members. Both were built on a newly unified recordkeeping platform and shaped by the Wealth and Health philosophy Transamerica launched in 2018, which treated financial and physical wellbeing as inseparable. The result was a set of experiences where a participant could check their retirement outlook, adjust contributions, and access financial education in one place, and an individual policyholder could view policy details and run calculators without picking up the phone.
→ Wealth + Health Brand Launch
→ Digital Platform Consolidation
→ Individual Experience
→ Participant Experience
PR Newswire
Asia Pacific Customer Portal
→ 2015 / 2017
I designed the Asia Pacific Customer Portal for PR Newswire, a self-service platform that let corporate communications teams across Asia submit press releases, track their orders, and measure distribution results in one place. Before this portal existed, clients sent their content and assets by email, waited on account managers for every status update, and had no direct view into how their releases performed after distribution. The process was slow, hard to scale, and invisible to the people doing the work. I came in after an outside agency failed to deliver, taking on architecture, UI design, and front-end development as the sole designer on a distributed team spanning Hong Kong, New York, Toronto, and New Delhi. I built the portal in three languages, navigated restrictions on Google-hosted services in mainland China, and shipped a product that reached 73% customer adoption across the region within its first year.
→ Press release submission
→ Order status tracking
→ Media asset library
→ Distribution channel selection
→ Post-distribution reports
→ Multilingual interface
→ Mobile-responsive email