← Back

Work

Selected projects, 2015 – present

Founder · Arizona Copper State Solutions Inc

AI Consulting

Through my consultancy I build working software for small and mid-sized operators — quoting tools, field apps, scheduling, AI features, automations, and the marketing around them. I use AI-assisted (“vibe”) coding to move at startup speed, and a product manager’s discipline to make sure it holds up in production.

Client names are withheld. Screenshots use sample data or have names, logos, and contact details redacted.

Live in productionWorking prototypeClient proposal

Applications & features

Application · Sales tooling

Dealer quoting platform with a 3D building preview

Live in production

Steel-building dealers were pricing custom buildings from paper flyers and spreadsheets — slow, inconsistent across states, and easy to get wrong.

  1. Pick size & style
  2. Price updates live
  3. See it in 3D
  4. Export order form + customer quote
  • State-specific price books with snow/wind engineering loads, tax, deposit, and an all-in estimate
  • Paste rough call notes and AI fills the quote; upload a price flyer and AI extracts the price tables
  • Saved quote history, side-by-side compare, shareable quote links, and PDF order forms

Distinctive moveBuilt around the dealer's real paperwork: outputs match the manufacturer order form, so a quote turns into an order without retyping.

3D steel-building configurator with size, roof, and color controls and a live starting price
Public 3D preview — change dimensions, roof pitch, and colors in the browser.
Quote calculator showing state pricing, building type, dimensions, and an itemized build estimate
Sign-in-gated calculator used by dealer reps.

Application · Field operations

Field-service notebook for a mobile mechanic & welding shop

Live in production

A one-truck repair business kept equipment history, parts orders, and quotes across texts, notebooks, and memory — in places with no signal.

  1. Log the machine
  2. Triage the fault
  3. Track parts on the way
  4. Draft the quote
  5. Export to the books
  • Seven modules — equipment, service logs, remote triage, parts cross-reference, parts tracking, quotes, route planning
  • Server-computed quote totals that always stay drafts; deterministic CSV hand-off to accounting
  • Installable phone app with honest offline state, private photo/PDF attachments, and global search

Distinctive moveZero-trust from day one: Cloudflare Access is the login, every record is account-scoped, and attachments never touch a public bucket.

Three phone screens: a dashboard of actionable counts, an equipment list, and a draft quote
Shown with sample data.

Application · Healthcare staffing

Shift scheduling for a home-care agency

Working prototype

An in-home care agency filled nurse shifts through group texts — no single view of who covers which patient, and no audit trail of approvals.

  1. Admin posts a shift
  2. Nurses get a text
  3. A nurse claims it
  4. Admin approves in one tap
  5. Coverage calendar updates
  • Phone-number-plus-code login for nurses — no passwords to forget
  • Each nurse sees only their own shifts; admins see the whole month and the approve queue
  • Patient coverage timeline and an SOS board for last-minute call-outs

Distinctive movePrototype first, so the owner could tap through the real flow before a line of backend was written — then rebuilt as a production app on Cloudflare Workers + D1.

Three phone screens: admin dashboard with an approval queue, a nurse's schedule, and a master coverage calendar
Interactive preview with sample data — no real patients or staff.

Feature · AI imagery

"See your home lit" AI visualizer with an instant ballpark

Live in production

Holiday-lighting buyers can't picture the result, and the installer spent evenings driving out to give quotes that often went nowhere.

  1. Upload a photo of the house
  2. Pick a light style
  3. AI renders the home lit
  4. Get a ballpark and book
  • Nine popularity-ordered styles, switchable after the render without re-uploading
  • Choose what to light — roofline, palms, trees, stakes, garage — plus free-text requests
  • Itemized ballpark range built from the installer's real per-foot and per-tree rates

Distinctive moveThe estimate is deterministic, not generated: rates were validated against real past jobs so the AI picture never promises a price the owner won't honor.

Style picker with nine light-color options and checkboxes for which parts of the home to light

Application · Logistics

Offline-first shuttle operations app

Working prototype

A truck-and-trailer shuttle service runs on remote rivers where there's no cell signal — bookings and drop-off locations lived in phone calls.

  1. Customer books
  2. Driver runs the day
  3. Photo at drop-off
  4. Operator steps in only on exceptions
  • One app, three roles — customer, driver, operator — switchable with tabs
  • Works in airplane mode; status updates and photos sync when signal returns
  • Installable to the home screen like a native app

Distinctive moveDesigned for zero bars first: the demo is meant to be tried with airplane mode on.

Phone screen of an offline-first shuttle app showing customer, driver, and operator roles
Sample data; texts are simulated.

Metal-panel takeoff estimator

Working prototype

Reads vector CAD elevation PDFs, lays out panels around every opening, and outputs dimensioned plots, cut lists, and CSV/JSON — tested line-by-line against the fabricator's own reference reports. No screenshot: the source drawings are confidential.

Branded presentation generator

Live in production

Upload a company deck or PDF; it extracts colors, fonts, logo, and disclaimer, researches the topic, and builds a new on-brand PowerPoint with charts.

Websites

Website · Local service business

Small-business sites built to generate calls

Live in production

Owner-operators need a site that looks established, ranks locally, and turns a visit into a phone call or quote request — without a monthly agency retainer.

  • Fast static or edge-hosted builds on Cloudflare and Vercel, with real job photos and verbatim reviews
  • Lead forms with spam honeypots, routed to email and SMS, and checked daily by synthetic monitoring
  • Service-area pages, local SEO structure, and DNS/email deliverability set up without breaking existing mail

Distinctive moveConcept sites are pitched before they’re sold: owners click through a working version of their own business, then we iterate on their feedback.

Homepage of a professional-organizing business with a call button and a photo of an organized closet
Business name and phone number redacted.

Automations

Lead routing that can't get lost

Live in production
  1. Web form
  2. Spam honeypot
  3. Email to owner
  4. Text to owner's phone

Every quote request lands in the inbox and as an SMS within seconds, with reply-to set to the customer.

Synthetic site & form monitoring

Live in production
  1. Hourly uptime + content check
  2. Daily honeypot form POST
  3. Alert email
  4. GitHub issue

Every client site and lead form is exercised automatically — without sending a fake lead — so I hear about a break before the client does.

Client-context indexing

Live in production
  1. Message thread
  2. Screen recordings
  3. AI transcription
  4. Searchable brief

Feedback that arrives as texts and screen recordings becomes a searchable index, so a client never has to repeat a request.

AI document extraction

Live in production
  1. Price-flyer PDF
  2. Vision model
  3. Structured price table
  4. Reviewed, then live

Manufacturer price flyers become structured pricing data instead of hours of hand entry — with a human review step before anything goes live.

Marketing support

Marketing · Paid social

Problem-first ad creative and capped 14-day tests

Client proposal

Small contractors burn money on boosted posts. I package ad creative, targeting, and a hard-capped test plan the owner can approve ad by ad.

  • Ad copy that opens on the customer’s problem, paired with real job photos or clearly labeled AI renders
  • Small capped budgets, a day-7 check, and scale-or-kill at day 14 on cost per lead
  • Instant-form leads routed to the owner by text so they get called back within minutes

Distinctive moveEach ad has its own Approve / Request change control, so the owner signs off on exactly what runs.

Three Facebook ad mockups for a painting contractor, each with approve and request-change buttons
Proposal mockups; engagement counts removed.
A 14-day ad test plan with a capped budget, market split, expected-lead range, and timeline
Lead range is a benchmark-based projection, not a reported result.

How I build

Requirements before code

Every client ask becomes a Given/When/Then requirement; every commit cites it.

Parallel AI agents

Multiple coding agents work isolated git worktrees at once; I review and merge.

Verified on the live site

Done means proven on the production URL — smoke tests and screenshots, not a green build.

Security defaults

Platform auth (never hand-rolled), secrets in 1Password, credential scans before every commit.

Model routing — the right model for each job

Live research — regulations, jurisdictions, current docs
Perplexity Sonar Pro
Structured reasoning, quote parsing, document extraction
Claude Sonnet
Photo-real image rendering and video transcription
Google Gemini
Authoritative data when an API exists
Direct APIs, not web search

Product leadership

Senior Product Manager

McDermott Will & Emery

  • AI platform serving 1,700+ attorneys and staff at a top-10 global law firm
  • Reduced research time and surfaced institutional knowledge across practice groups
  • Drove adoption strategy across offices in the US, Europe, and Asia
McDermott Will & Emery platform screenshot 1McDermott Will & Emery platform screenshot 2

Product Manager, R&D

7-Eleven

  • Owned product for computer vision self-checkout in the R&D division
  • Partnered with hardware and ML teams on food automation robotics
  • Moved technology from lab environment to pilot store deployment
7-Eleven computer vision self-checkout7-Eleven food automation robotics

Senior Manager II, Product

Walmart US

  • Built field tools deployed across 4,500+ stores for the #1 Fortune 500 company
  • Mobile-first workflows for asset management, incident response, and preventive maintenance
  • Coordinated across facilities, engineering, and operations at national scale
Walmart field tools screenshot 1Walmart field tools screenshot 2

Product Manager

COPLEX

  • Launched 50+ companies from concept to first revenue at a Phoenix venture studio
  • Owned discovery, MVP scoping, and go-to-market across each engagement
  • Worked across fintech, healthtech, enterprise SaaS, and consumer products
COPLEX venture studio project 1COPLEX venture studio project 2COPLEX venture studio project 3

Education & Other

Arizona State University

Ira A. Fulton Schools of Engineering

Thunderbird School

Global field study — Colombia, Ecuador, Chile

altMBA

Seth Godin's leadership program

Habitat for Humanity

Build teams — Nepal, Tanzania, Sri Lanka, Romania

Connect on LinkedIn →