Projects

Things I've built. Each one started with a problem worth solving.

Ripple Treasury Platform

Redesigning navigation for a platform that processes billions in daily treasury operations.

Problem

The platform’s navigation was the top reason deals were lost in competitive evaluations. Users across 160 countries struggled with a structure that hadn’t been meaningfully updated in 8+ years.

Solution

Led a research-driven redesign: hundreds of user quotes analyzed, data-driven personas, 11 navigation patterns prototyped, and card sort studies with participants across the organization. Introduced a task-oriented framework replacing technical module names with solution-based categories.

Key Metrics

160+ countries11 patterns prototypedFirst redesign in 8+ yearsResearch-driven

Stack

React, TypeScript, Vite, Pendo Analytics, Figma

Design Token System

Multi-brand design tokens powering two product lines from a single source.

Problem

After Ripple acquired GTreasury, two brands needed to coexist in one platform. Manual theming was error-prone and couldn’t scale across product lines.

Solution

Built a token pipeline that outputs to 6 formats from a single YAML source. Supports multi-brand architecture with semantic aliasing — GTreasury and Ripple branding from one system, deployed automatically through CI/CD.

Key Metrics

6 output formats (CSS, Tailwind, TS, JSON, SCSS, LESS)2 brands, 1 sourceAutomated CI/CD distribution

Stack

Style Dictionary, YAML, design tokens, CI/CD pipeline

Analytics Practice

Building the analytics muscle for an enterprise treasury platform.

Problem

The product team was making decisions without data. No segmentation strategy, no adoption tracking, no structured feedback loop across tens of thousands of monthly platform visitors.

Solution

Established a Center of Excellence with a governance charter, built a 3-tier adoption framework, created segmentation strategy across products, and connected analytics to product decisions through executive dashboards and automated reporting.

Key Metrics

Enterprise-wide analytics coverageComprehensive account segmentation3-tier adoption frameworkCenter of Excellence established

Stack

Pendo, Python, SQLite, Power BI

Nodus

AI infrastructure that turns a chat window into a chief of staff. Built right. Left open.

Problem

The most powerful way to work with AI is locked behind a technical wall. Knowledge workers use AI as a chat assistant — copying, pasting, starting every session from zero. The leap to persistent context, agent orchestration, and trustworthy output requires infrastructure that nobody had laid.

Solution

Nodus is the infrastructure layer between you and AI: structured knowledge management (630K+ vectorized chunks, 5ms search), 17 specialized agents coordinated through a context bus, trust architecture with provenance on every output, and full observability. One dependency. Self-hosted. Local-first. No data leaves the perimeter.

Key Metrics

1 dependency (pyyaml)17 specialized agents630K+ searchable chunksLocal-first architecture

Stack

Python, FastAPI, SQLite, sentence-transformers, Phoenix observability

Structured Thinking

Visit →

A publication on why design needs what finance already built.

Problem

Design knowledge is trapped in tribal knowledge and scattered style guides. Finance solved this decades ago — FIBO has 1,346 classes and 273 object properties. Design has JSON tokens with string values.

Solution

Structured Thinking explores building the semantic layer above W3C Design Tokens — formal ontology that lets both humans and AI reason about design decisions, not just exchange variables.

Key Metrics

W3C DTCG extensionCross-discipline (finance + design)

Stack

Substack, research, ontology design