Design · Developer · One-man studio

I design it, build it,and keep it running.

UI/UX for Toyota Research Institute, Cisco, Sysdig, John Deere and SAIC Motors — then I learned the rest of the stack. Ratchet AI: idea to App Store in about thirty days, solo.

The pipeline · 1 of 6 opened

Follow the dashed path 01 → 06. Click a stage label on the sphere, a sector, or a layer here.

Active products · in production

00

Fig. 01 — geodesic index · six sectorsSEC 01 · LAT +62.0 · LON +0.0 · CELLS 61
Layer 01

Research & problem framing

Research & problem framing
Toyota Research Institute — AI UX research

Fourteen years of this part: user journeys, stakeholder pain points, UX metrics with baselines. It is also where I kill my own ideas — Ratchet started as a post-modern sci-fi tuner interface, and the honest answer was that it needed to read as an AI app instead.

User journeysUX metricsStakeholder mappingDesign sprints

14 yrs

UI/UX practice

~30 days

Ratchet AI, idea to App Store

40+

Live OBD2 PIDs in Tuner Mode

1

People between you and the code

The register

Every product, its state, its stack
01

Ratchet AI

On the App Store
Ratchet AI

An AI mechanic. React Native iOS/Android app that pairs an OBD2 Bluetooth adapter with tier-based LLM diagnostics — DeepSeek V3 and Claude, voice interaction, hard safety gates on brakes and airbags, and a Tuner Mode logging 40+ live PIDs.

React NativeNode.jsStripeRevenueCatPostHog
tryratchet.com →
02

Woven Platform

Running, not a concept
Woven Platform

An AI-assisted 3D pipeline and world builder for the open web. Reference image → Trellis2 → automated Blender post-process → Draco compression → web-ready GLB, loaded straight into a browser world builder. No install, no ads, no dark patterns.

Babylon.jsPlayCanvasThree.jsComfyUIBlender
wovengames.app →
03

Lasso AI

In build

Send me the one-liner and the URL and this fills in. Placeholder copy stays off the page until you do.

AI WorkflowFull Stack
Copy + URL pending →
04

Paladin

In build

Same — one line on what it does, where it runs, and what it's built on, and it joins the register properly.

Full StackProduction
Copy + URL pending →
Study · Woven Platform

Image in. Game asset out.

I was a 3D modeler at Disney Interactive a lifetime ago, so I can still poly-model — the question was whether an image-to-3D pipeline could carry a one-man studio. It can, with rules. Trellis2 infers back-face geometry from a single reference and stylizes faces convincingly; it also fails on interiors, tunnels and anything large with complex detail, and it won't optimize geometry the way a game artist would. So the pipeline is bounded to what it's actually good at: props, set dressing, and mood.

Works

Props and objects from a clean reference. Inferred back geometry that stays true to the design. Stylized characters in a relaxed or T pose.

Fails

Interiors and tunnels generated from inside. Text on surfaces. Large buildings, even at high polycount. Decimation with complex detail present.

Fix

Prompt the source image for 3D generation, not for looks. Go high-poly then compress with Draco rather than asking the model to go low. Hand-model the hero assets.

Concept art — the dying city
Concept — the slums district
Generated prop
Prop — image to GLB, what works
Rigging
Rig — the part still done by hand
Woven toolchain · in-browser editors
[ Mesh Sculptor — In Build ]
A

Mesh Sculptor

Screen grab pending.

[ Mesh Painter — In Build ]
B

Mesh Painter

Screen grab pending.

[ Shaderade — In Build ]
C

Shaderade

Screen grab pending.

Track record

Where the fourteen years went
2026 —Ratchet AI / Starfox IndustriesFounder & designer-engineer. Designed and shipped the whole product solo — app, backend, store submission, payments, analytics.
2025 – 2026Toyota Research InstituteAI UX designer. LLM-to-CAD, aerodynamic body optimization, AI ECU chatbots, distracted-driving simulation, EV battery discharge visualization — and the TRI design pattern library.
2024 —Starlyte ImagingChief designer. Confocal microscopy workflows mapped end to end, a component library for Python development, hardware-control prototypes, plus the brand and the website.
2021 – 2022AlkiraUX engineer, autonomous networking. Owned the design system library and shipped end-to-end features on it, including the policy-impact view across the CXP.
2019 – 2021Sysdig · SpiketrapContainer and Kubernetes security UX — build and runtime vulnerability scanning against policy and compliance. Then LLM sentiment analysis at Spiketrap, later acquired by Reddit.
2016 – 2019SAIC Motors · Ford · Health 2047Led four designers on an autonomous ride-hail HMI for the Maxus flagship. Motion design for F-150 and next-generation clusters at Ford. HIPAA-compliant physician-patient app at the AMA's Health 2047.
2008 – 2013Disney Interactive · Nexon MGame artist, then senior artist and technical lead: concept art, 3D models, UVs, rigging, animation and UI, and the Unity pipeline behind several titles. MFA, Academy of Art.
Toyota Research
Toyota Research
Starlyte
Starlyte
Sysdig
Sysdig
John Deere
John Deere
SAIC Nexus
SAIC Nexus

Full case studies live on marktsedesign.com — the archive stays up.

Archive · 2008 – 2013 Works

marktse.net 2010

Early 3D models, concept art & UI systems from Disney Interactive, Nexon M, and indie game prototypes. Preserved directly from the original legacy portfolio.

3D Brute Character
2008 · 3D
Exhibit 3dbrute12008

3D Brute Character

Full body sculpt and high-to-low poly modeling in 3ds Max & ZBrush.

Select artwork (33 items)
How to engage

Design to dev, complete

A business problem in, a shipped application out. Research, visual design, app, backend, store submission, payments, analytics. Same scope I run on my own products.

Dev-as-a-service subscription

A monthly slot on my calendar instead of a project. Pause anytime, renew anytime. Best for founders who need continuous shipping, not a one-off handoff.

AI & 3D specialist work

LLM product surfaces, inference UX, data visualization, and web 3D pipelines. The work TRI, Sysdig, John Deere and OmniAb hired me for, now with the build attached.

Questions people actually ask

You're a designer. Can you really build it?

Ratchet AI is on the App Store: React Native app, Node backend, Stripe web payments, RevenueCat IAP, PostHog analytics. Designed, written and submitted by me in about thirty days.

Do you use AI to write the code?

Yes — Claude Code and Codex, in plan mode, with me reading every diff. You still have to be a designer first: AI kills the polish tax, not the thinking. Wrong flow built fast is still a wrong flow.

Who owns the code?

You do. Your repo, your cloud, your app-store accounts, your domain. Nothing proprietary of mine sits in the middle and nothing makes leaving expensive.

What happens if you get hit by a bus?

Everything is documented and standard — React Native, Node, Postgres, Vercel, Figma. A competent contractor can pick it up cold. That's a design constraint, not an afterthought.

Will you take on design-only work?

Rarely, and only where someone credible is building. Screens nobody shipped are why this studio exists.

How fast can you start?

One build runs at a time. Scoping calls this week; subscription slots open month to month.

Tell me what should exist.I'll tell you what it takes.

A paragraph is enough to start. You get a real answer from the person who would design it, build it and submit it — scope, rough number, and whether I'm the wrong choice.

Replies within one business day. Nothing gets added to a mailing list.