Applied AI · Operations · Project delivery

I turn operational problems into working systems.

I lead hospitality operations and build tools around workflows I know firsthand. My strength is defining the problem, directing AI-assisted implementation, and verifying that the result works in practice.

See the work View résumé

Function Spaces timeline: a week of events laid across rooms, with setup chips, guest counts, and a seating-diagram thumbnail.

Synthetic data

01 — Flagship · deployed

The operating layer Delphi didn’t provide.

Amadeus Delphi held the source data, but its exported event detail was too slow and fragmented for banquet leaders moving through a 30,000-square-foot operation. I rebuilt it as one view: rooms, assignments, diagrams, and status, on any browser.

~15: daily users
500+: automated tests
Live: in production

Next.js · TypeScript · Postgres · Playwright

See Also on iPhone: an article cover for Bioluminescence, with a full-bleed photograph above the title set in Cormorant.

02 — Native product · shipped

A calmer way to follow a Wikipedia rabbit hole.

A paid native iOS reader with nearby discovery, offline bookmarks, accessibility support, and five reading styles. I owned product definition, iterative review, App Store submission, and release operations.

5: reading styles
Paid: native iOS, shipped

SwiftUI · SwiftData · MapKit · CoreLocation

View on the App Store

The counted sheet: nine spirits with counts filled from speech, and one line held back for review where the parser was not confident.

Synthetic workbook

03 — Deployed POC · passcode-gated

Count the bar by talking to it.

A manager counting alone has to speak, remember, and transcribe at once. This listens, matches spoken items against the property’s own sheet, and surfaces only the low-confidence lines — then patches the counts back into the original workbook.

87.5%: first synthetic auto-fill
87: automated tests
2: ASR providers, one interface

AssemblyAI · Deepgram · Postgres pg_trgm · XLSX

iMessage
Queue
Agent
Tools

Every request takes the same path, and every step is logged

04 — Personal agent harness

Long-running agents, kept on a leash.

LAUTREC reconstructs conversation context across stateless invocations, suppresses superseded drafts before they send, and routes authorised requests into calendar, email, task, media, and development workflows.

Claude Code · Codex · Gemini · BlueBubbles

Training slide explaining a 1.25 ounce banquet pour standard.

05 — Training & enablement

Operational knowledge, made teachable.

A 31-slide bartender learning program with explicit objectives, demonstration, case studies, an assessment, and job aids. It turns unwritten expectations into a repeatable training experience.

31: slides, 7 sections
3: delivery modes

Lecture · Lab · Practicum

01of 05

How I work

AI-assisted does not mean unowned.

The tools accelerate implementation. Responsibility for the workflow, decisions, verification, and consequences stays with me.

01
Discover
Observe the real workflow, its users, exceptions, and cost of failure.
02
Specify
Define behavior, trust boundaries, acceptance criteria, and what stays deterministic.
03
Build
Direct agents, isolate work, review changes, and iterate against actual usage.
04
Verify
Use tests, logs, synthetic cases, human approval, and post-action checks.

What I personally own—and where AI assists

I own

Workflow discovery, requirements, system behavior, data rules, acceptance criteria, task decomposition, review, deployment decisions, and iteration with real users.

AI assists

Implementation, test generation, debugging, documentation, and code review under a supervisor-and-subagent workflow.

I verify

Deterministic parsing, role behavior, failure handling, UI workflows, and regression coverage against operational needs.

About

I came to technology from the floor, not the classroom.

I regularly lead or coordinate 20–40 people in live hospitality operations, where plans meet guests, deadlines, room turns, and imperfect information. That background shapes how I build: the system has to be legible, recoverable, and useful under pressure.

I'm completing a B.S. in Hospitality Management at Texas A&M University, graduating December 2026, and targeting applied AI, implementation, enablement, and operations-technology roles in Austin or remotely. Extensive travel is welcome.

GitHub ↗LinkedIn ↗Résumé ↗

Open to the right next step

Have a messy workflow worth fixing?

I'm interested in teams that value operational judgment, responsible AI adoption, and people who can translate between users and engineers.

brayden@toulousemail.com