Jeremy HutchcraftBrick City AI EngineerBook a call

Brick City AI Engineer · Ocala, FL

Anyone can prompt an app.Knowing when the AI is wrongis why you bring in the big guns.

I build your product with AI doing the typing and me making every call. My low prices reflect the efficiency with which I operate AI.

I'm Hutch. I vibe-code too. There's a time and a place for it. Code your customers touch is not it.

Book the free half hourSee what I catch

Things the AI told mewith a straight face.

All of these happened to me. Each one starts with what I'd asked for, in plain English, then what was really going on.

  1. A new feature was ready to go live. The AI ran the automatic checks that are supposed to prove software works.

    “All tests pass.”

    What was really going on

    Every check was green. I used the real thing anyway, the way a customer would, and found about fourteen more bugs. Your customers would have found them instead.

  2. Before we built on a set of records, I asked how complete they were.

    “99.96% of records are populated.”

    What was really going on

    It had counted the placeholder word "auto" as a real answer. The number was wrong, and it was already sitting in the write-up the team was going to read.

  3. I was researching how late payments get handled before a billing change.

    “There's a 72-hour grace period.”

    What was really going on

    There isn't one. It invented a billing rule and wrote it down like a fact. That is how a made-up promise ends up in a customer email.

  4. I asked for one small edit to a page customers read.

    “I only changed what you asked.”

    What was really going on

    It had also slipped in wording nobody approved. I read every change, especially from the cheaper, faster models.

  5. A kids' learning game was marking right answers wrong. I shipped a fix.

    “Fixed.”

    That one was me

    My fix recreated the same bug one screen over. It only hit kids typing on a keyboard, never tapping, so the usual check missed it. I caught it by checking what was actually live.

Don't take
my word for it.

Receipts.

  • OwnAmerica

    A rental investment marketplace. I was the full-stack engineer and UX lead.

    Acquired by Renters Warehouse, 2019

  • 64 Analytics

    College baseball and softball analytics. Somebody handed me a spreadsheet. I built the data pipeline, the web app and everything under it, and ran it through three NCAA seasons.

    Acquired by Rule42, October 2026

Two companies I built the product for. Both got bought.

The Match tab on a player page: 308 programs match the player's preferences, with one program ranked first at an 82 match, team stats below and a matching-criteria panel on the right. The player shown is fictional.

The Match

The player-and-coach matching system built for 64 Analytics Player Profiles to aid recruiting.

A player says what matters to them, and every program gets scored against it, with the reasons shown.

The player on screen is made up. The screen is real.

The score breakdown for one program: bars comparing the player's ranking with the team's in each category, a radar chart of fit across eight categories, and sliders for adjusting the matching criteria.
Why a program scored the way it did, and the sliders that change it.

Also built for 64 Analytics

Nothing gets lost, and nothing gets started half-baked

Every bug report and feature request lands in one place and gets sorted before anyone touches it: how urgent it is, how big it is, and whether it is actually ready to build. Work that is safe to hand to an AI agent goes to the agents. Work that needs a person goes to a person.

For the technical reader: an issue intake and triage system in GitHub. A needs-triage and needs-info intake, priority P1 to P4, estimates XS to XL, status from backlog through ready, implementing, testing and review, a ready-for-agent versus ready-for-human split, and routing across the client app, database, data-import and play-by-play pipelines.

A data pipeline that only costs money when it runs

The data comes in every night, the numbers refresh, and the right people get told. An hourly watchdog keeps an eye on it. No servers sit switched on waiting: the heavy machine only starts when there is work for it, so the bill follows the work.

For the technical reader: a Terraform-managed stack on AWS, built on Lambda, EC2, EventBridge and SQS. Lambdas handle the nightly data import, a post-import materialised-view refresh, notifications, S3 cleanup and player-profile file handling. An hourly snapshot watchdog reports CloudWatch metrics. A scheduled, EventBridge-driven play-by-play pipeline starts an EC2 worker only when there is work.

The day job

I built a test harness that checks whether an AI's skills actually fire, and a loop that turns a reviewer's comment into a pull request. Half my time now goes to teaching a whole engineering team to do this safely.

Around the house

A finance system that reads the bank and texts me every morning. A spelling and math game for my kid. I build the same way at home.

$90an hour.

Same for everybody. Posted here so nobody has to sit through a sales call to learn a price. The first half hour is free, and no job is too small.

  • Smallest start

    Ten hours. I jump into your next sprint and get something shipped.

    $900

  • Second opinion

    I read your AI-built code and tell you what is wrong with it. About ten hours.

    about $900

  • Blueprint

    Spec, architecture and a number you can plan around. About twelve hours.

    about $1,080

  • The build

    Itemized, every hour accounted for, billed on your sprint cadence.

    $90 an hour

  • Handoff

    Your team learns the workflow and gets the keys. About sixteen hours.

    about $1,440

  • Fuel

    Ten, twenty or forty hours a month after launch, for fixes, features and hard calls.

    $675 / $1,350 / $2,700

  • Back-to-back sprints

    Every sprint in a row after your first.

    25% off

How I keep the AI honest

  • Types and unit tests passing is necessary. It is not sufficient.
  • Verify as you prompt. Prune as you prompt.
  • The scratchpad is not the document.
  • Five cheap narrow agents beat one long tired thread.
  • Nothing ships until a person who can defend it says so.

Send me your repo.
Or your napkin sketch.

Book the free half hour

I'll tell you straight whether you need me. Every project I take on becomes my baby, so I'm picky about saying yes.

AI recommendations and workflow support are not legal, financial, clinical, cybersecurity, or compliance advice. High-risk workflows may require review by qualified professionals.

© 2026 Jeremy Hutchcraft · Brick City AI EngineerOcala, FL