Caolan McArthur

AI systems and evaluation · games · operations since 2003

Cheshire County, New Hampshire cato@catocaolan.com Available now

Now

I build and evaluate agent systems, and I ship things. Most recently thirteen games — designed, built and released end to end, with an agent fleet and no coding background when I started.

I came to AI systems engineering out of operations, quality assurance and management going back to 2003: manufacturing lines, hotels, film, retail, food service. That background is the reason I'm useful here rather than a detour on the way. Evaluating whether a system actually does what it claims is the same discipline whether the output is a moulded part, a hotel night audit, or an agent's answer.

Taking short contracts in agent evaluation.

Thirteen games, shipped

Built for MapChat and released on the iOS App Store in 2026. Designed, built and shipped end to end. Every title is public and countable.

  • Imprompto
  • Emoji Off
  • Pair Play
  • Trivia Night
  • Crowd Oracle
  • Spill It Out
  • Blackjack
  • Spades
  • Poker Texas Hold’em
  • Gin Rummy
  • Whist
  • Hearts
  • Same Page

Built with an agent fleet of seventeen named seats — a floor, not a total; each seat could dispatch workers beneath it and those were never counted. Two colleagues oriented me on databases, React Native integration and serverless edge functions. I took that and built the catalogue.

Areas of practice

  • Agent systemsDesigned, dispatched and coordinated a fleet of seventeen named agent seats to build and ship a thirteen-title commercial catalogue end to end — from concept to live release on the App Store.
  • Evaluation and quality assuranceTwo industries in formal QA — manufacturing and games — plus verification threads through hotel night audit. In-process inspection, formal sampling against spec, adversarial review, and the discipline of not trusting an instrument until you've checked what it can't see.
  • Game design and directionDesigned and directed a commercial video-game catalogue end to end, from concept to live release.
  • Operations and managementContinuously since 2003. Repeatedly hired at the bottom and promoted; teams from four to twenty-five; both ends of the power dynamic, often simultaneously.
  • TechnicalDatabases · React Native integration · serverless edge functions · App Store submission.

Work history

Newest first. Back to 2003, and that is a floor — work before then was informal. Roles marked ran concurrently with the job above them, so they sit under it rather than in date order.

RoleWhereWhen
Head of Growth and GamesMapChatNov 2025 – Jul 2026
Quality Assurance TesterDire Wolf DigitalMay 2021 – Mar 2025
Co-founderRalitan CreativeOct 2018 – May 2021
DriverUber and Lyftc. 2014 – 2020
Assistant DirectorCinemaRavenOct 2014 – Oct 2018
Five positionsHotel BoulderadoAug 2010 – Oct 2018
Night AuditorHamptonDec 2011 – Mar 2012
Night SupervisorDel TacoMay 2010 – Aug 2010
CanvasserGrassroots CampaignsApr 2010 – May 2010
Sushi ChefThe Sushi BarOct 2008 – Mar 2010
Jomar Operator – QAHCP Packaging USAMar 2007 – Nov 2007
SupervisorGameStopSep 2004 – Jan 2007
Integrated Product TeamBordersNov 2005 – Jul 2006
CloserSubwayJul 2003 – Aug 2004

Hotel Boulderado — five positions across eight years. Entered on night audit and rose to Front Desk Manager: night audit → front desk → Front Desk Manager → Bell Captain → Front Desk Manager. The order is confirmed; the boundaries between titles are not, so this prints the order and not the years.

Through-lines

Hired at the bottom, promoted fast

The pattern repeats across two decades and most of the employers above: entry-level hire, rapid promotion into management. It is also my answer to the degree question — the record shows what happens after someone lets me in the door.

Two and three jobs at once, for years

I drove rideshare concurrently with Hotel Boulderado and CinemaRaven for roughly four years — three jobs at a time, and both ends of the power dynamic at once: managing staff by day, driving for an algorithm at night.

Quality assurance, twice over

Formal QA in two industries. The first was a manufacturing line in 2007 — continuous in-process inspection of my own output plus a formal sampling pass against spec twice per shift. The second was four years testing games. A third verification thread runs through hotel night audit at two properties.

The management arc starts long before the hotel years

Retail supervision from 2004 — six years before the hotel. The management record runs longer, and starts earlier, than a résumé arranged by job title suggests.

Education

Autodidact.

Contact

cato@catocaolan.com

Available now for short contracts in agent evaluation, and for longer work.