SLMMDA / INDEX 01 — HOME
OKLAHOMA CITY · 35.47°N 97.52°W
AI PRODUCT
DESIGNER &
BUILDER
STUART MARTIN — PORTFOLIO 2026
I turn ambiguous problems into clear working software — combining product strategy, UX architecture, visual design, and AI-assisted development.
AVAILABLE NOW
RÉSUMÉ
SEE THE WORK
SCROLL
One product. Four months. Three business demos. Five stop decisions. Twelve years of design underneath it.
682
commits — one product, four months, sole author
3
businesses — real demonstrations, real outcomes
5
stop decisions — projects ended on evidence
12
years of design practice underneath all of it
The outcomes live on the project pages.
01
Flagship
Kip — Built · Demonstrated · No paying customers
I built a local-first AI workspace, took it to real businesses, and the companies that liked it still said no.
Kip keeps company knowledge on-device, answers with citations, and previews work before anything runs. The harder problem turned out to be adoption.
Kip — pipeline, pending approval
KIP / PIPELINE — INTERACTIVE DEMO PLACEHOLDER
A pending plan shows the literal email, meeting and task before approval.
02
Flagship
Gemmark — Delivered · Paused before production
I was hired to build a brochure site. Discovery found a commerce platform — and an audit told me not to rush into building it.
18 architecture documents. 37 approval records. 15 cross-item conflicts resolved before code.
Read the case study
GEMMARK / DISCOVERY DIAGRAM — PLACEHOLDER
03
Also
Six more projects, three of which I stopped
The flagship projects go deepest. Six more show what I build, test, change, and stop.
Different domains. Same pattern: define the system, make the evidence visible, then act on what it says.
/ 01
Beepo Hub
A content-automation engine, verified end to end with 1,219 passing tests and a design system authored as a machine-readable contract. Two defect discoveries in it are the best evidence I have that I read systems rather than symptoms.
/ 02
Rally
A household task app that modelled chore verification as its own state machine. I stopped it after working out that the reward economy it needed to grow would make verification adversarial.
/ 03
PolyLab
I measured 1,096,034 rows of a prediction market’s tape, proved the strategy unback-testable in principle, and stopped in a day. A negative result that cost one day and saved several hundred hours.
/ 04
Lead OS
Built and documented my own company’s lead-to-proposal pipeline — first contact to a Monday summary, no manual entry.
/ 05
CRE Discovery System
Interviewed a broker, turned his criteria into a scored alerting system — then stopped it after reading the source platforms' terms about automated scraping.
/ 06
Story Studio
waiting written brief.
See all work
04
How I work
Method
I don’t type most of the code. I design the system that tells an AI implementer what to build, how to prove it, and when to stop.
35 specialised reviewers. 24 reusable procedures. 12 operating rules. One tested recovery system.
METHOD / GOVERNANCE DIAGRAM — PLACEHOLDER
05
Underneath
Twelve years of design practice
Before product design, twelve years of graphic design taught me to make systems that survive the medium.
Brand, packaging, editorial, typography and print are not the old work. They are the craft underneath the product work.
Given a one-off, I still look for the reusable system behind it.
See the archive
FOUNDATION / ARCHIVE MEDIA — PLACEHOLDER
06
Next
Contact
I’m looking for the kind of product problem that is still blurry when I arrive.
Founding and early product teams are where discovery, architecture, interface and implementation direction still belong close together.
Based in Oklahoma City. Open to remote.
Get in touch
Résumé