Work

The work

Real projects and real numbers, so you can see what a fix looks like before you hire me. Two client projects, then the enterprise record behind them. Numbers are rounded down.

Client · Knoxville, TN · 6 days

McKee Outdoors

A fishing tackle shop in Knoxville, TN. Its website sat unfinished for 2+ years. It was finished in 6 days, Sept 28 – Oct 3, 2026, using AI-based workflows I designed and built for the job.

The results

  • Online store2,000+ productsLive online, up from about 500.
  • Sizes & colors23,000+Up from about 1,600.
  • Register sync90%+ of the storeSynced with the register within minutes.
  • New products1,200+Brand-new products, researched and added.

Read the full McKee Outdoors story

Installed · in hidden testing on both websites

A regional cemetery organization

Two cemeteries, one office. Name kept private. An AI assistant on both websites that answers families' everyday questions at any hour, and knows exactly when to hand a family to a person.

The problem

Families call with the same questions, often outside office hours, and often in the days after a loss.

  • Plots
  • Services
  • Directions
  • Hours
  • Decorations
  • Faith teaching

Why this one needed rules, not just AI

A cemetery is the wrong place for a chatbot that guesses. A wrong price, a promise about a plot, or a careless answer about faith teaching to someone who just lost a parent does real harm. So the assistant runs inside rules I built in code, which the AI can't talk its way around:

  1. It answers only from the cemetery's own approved knowledge base, and every answer has to cite its source. No source, no answer.
  2. A filter written in code, not the AI, blocks dollar figures, promises, and claims about who is buried where before anything reaches the screen.
  3. Three separate checks watch for grief, a recent death, pricing and faith teaching. If any one fires, the family gets the office phone number and a person, not a paragraph.
  4. Faith teaching is held back until the cemetery's clergy reviewer approves it, section by section, before it can be used.
  5. It says it's an AI in its first message, every conversation is logged for staff to review, and staff approve every change to what it knows.
  6. A 110-question test suite, written with the cemetery and including bereavement questions, runs after every change. Every miss in testing becomes a permanent test.

Timeline

  • Sep 9SignedFixed price, 12-week plan: build, red-team, supervised beta, launch.
  • Sep 14 – Oct 2BuiltKnowledge base from the cemetery's rules, website and faith documents. Safety rules, the assistant, and a staff tool to edit and approve answers.
  • Oct 7104 of 104First full live test run: 99 of 104. Fixed and re-run to 104 of 104 the same day.
  • Oct 7 – 12Clergy approvedThe cemetery's clergy reviewer went through every faith-teaching section; only approved sections can be used.
  • Oct 14Named and connectedThe client named the assistant. Call-back requests reach staff by email within a minute.
  • Late OctRed-teamedHard and hostile questions on purpose. Each gap became a new rule or test.
  • Nov 10InstalledLive on both cemetery websites in hidden test mode, with an in-chat feedback form for staff.

What's next

It goes public on both websites once the cemetery's attorney approves the privacy policy. Everything runs on accounts the cemetery owns, so at hand-off I hold no keys.

Before Easier Mondays · 8+ years

The enterprise record

Results from 8+ years running and improving IT operations at a large wireless carrier. Each number comes with what it meant for the people involved.

  • Caller verification$300K+ a yearSaved by a caller-verification tool I built.What that means: Callers were verified automatically, so less time and money went to manual checks.
  • Network AIIn productionAn AI tool I built that spots service-impacting network problems. It won an internal hackathon.What that means: Problems affecting customers got noticed sooner.
  • API monitoring85% less timeSpent validating changes, after I built API monitoring.What that means: Checking that a system change worked took a fraction of the time.
  • Recovery dashboards20% betterYear over year, in time to recover from outages, driven by dashboards I built.What that means: Outages got fixed faster, year after year, because everyone could see the numbers.

Recognized with a Continuous Improvement Award. ITIL v3 Foundation certified.

Next step

Your business could be next on the list.

Start with a free 30-minute call. You don't need to know what to fix.

Let's talk