Portfolio
Working resources, public code, real screenshots, and private-system case studies—each labeled so you know what evidence is available.
Based in the SF Bay Area and working remotely. Reach me at contact@sonomasolutions.io.
Higher education & accessibility
Course-data preservation and migration preparation
Preserved 23,760 course-version records and 23,760 source PDFs outside a legacy portal.
See how this work fits alongside accessible web systems and document remediation on the higher education services page.
Document remediation system
A multi-critic document remediation engine in Python, with audit-logged quality-assurance loops.
Private system, no public URL
Accessibility reference build
More than 60 paired accessibility reference pages in a custom WordPress library.
Screenshot case study
Full-stack AI Engineer
Live, open it yourselfOur flagship AI intake assistant. Use it yourself on a demo business.
Sonoma Solutions' flagship AI intake assistant. There is a working demo you can use: it runs a fictional Sonoma County HVAC company, answers questions from that company's knowledge base, collects the details a real job would need, refuses anything outside its rules, and hands back the summary the business would receive. The assistant at emissary.sonomasolutions.io answers questions about Emissary itself instead. Astro 4, React, and TypeScript on the frontend; Node and TypeScript backend modules for database access, a knowledge base, intake email parsing, lead routing, and downstream email integration. It is the productized version of the AI intake systems we build for clients.
Knowledge architect, AI engineer
Screenshot case studyA structured knowledge base of evidence-tagged atoms, turned into an exhaustive client reference manual.
An exhaustive, client-ready reference manual for MarketBrew (a deep SEO analytics platform), generated from a structured knowledge base. We extracted MarketBrew’s public reference material into evidence-tagged "knowledge atoms" through a parallel multi-model pipeline, then built the manual from that vetted database under strict schema and QA gates.
AI Engineer
Private system, no public URLVideo knowledge extraction with a real eval harness. Node.js + TypeScript on the Anthropic API.
A Node.js and TypeScript pipeline on the Anthropic API for video knowledge extraction. JSON Schema defines the knowledge atom shape with evidence validation on every persisted record. Real eval harness with eval-retrieval-multi, schema-conformance QA, and baseline snapshot scripts for regression detection.
AI Engineer
Private system, no public URLAn eval-first agent on top of the MarketBrew knowledge base, with answers cited back to the atoms.
The agent layer built on top of the MarketBrew knowledge base (see the Reference Manual project). An eval-first agentic scaffold on PostgreSQL: contracts and fixtures are defined before generation, and every recommendation is evidence-cited back to the underlying knowledge atoms. The scaffold pattern transfers across client domains.
Automation Engineer
Private system, no public URLEvent-driven email workflow in production. Inbox triage, classification, routing, human review.
Event-driven distributed email workflow in production. n8n with custom code handles inbox triage, classification, routing by topic and priority, structured event logging, follow-up scheduling, and human-in-the-loop review for high-priority replies before they go out.
Brand, copy, design, build
Live, open it yourselfA complete campaign brand and site for the AI-search work, built end to end.
A standalone brand and campaign site for the AI-search visibility work, built so the measurement offer could be sold under a name of its own rather than buried inside a general agency site. Five pages on Astro and Tailwind: the pitch, two fixed-price offer pages, a personalised prospect demo, and a held placeholder. Answer engine optimization is baked into the build rather than added afterwards, which is the least a site selling AEO can do.
Data engineer, AI engineer
Live, open it yourselfThe instrument behind the search work. Public code, and a sample report you can run with no API key.
A Python measurement harness that audits how often a brand and its competitors get named when buyers ask ChatGPT, Perplexity, or Gemini a question in their category. Config-driven from YAML, so swapping the query set and the competitor list points it at any market. It is the instrument behind the search-visibility work: rather than asserting that AI answers matter, it measures what the answers actually say, repeatedly, and shows the runs.
Developer
Live, open it yourselfSixty-two flashcards in a single HTML file. No build, no dependencies, no tracking.
A small, self-contained study app for the Utah driver license knowledge test: 62 flashcards covering BAC limits, DUI penalties, signs, and right-of-way, with a practice quiz and a condensed cheat sheet. One HTML file, no build step, no dependencies, no tracking. It loads instantly on a phone and works offline once opened, which is the whole point when you are revising in a queue at the DMV.
Designer, developer
Screenshot case studyA conversion-focused local service site on Astro, React, and Tailwind.
A fast, conversion-focused marketing site for an owner-operated pool and spa service in South Orange County. Built on Astro, React, and Tailwind to turn local searches into quote requests, with a multi-step contact form and per-city service-area pages.
