My philosophy on design

Techequity AI - UI/UX design (2026)

We host 60+ AI events and bring together 2,000+ people each year. The challenge I faced was helping visitors find what they needed before they left the site. I worked on clearer navigation, simpler content, and more obvious next steps.

The 2026 event page. Our events gather engineers, founders, policymakers and researchers, so I used colour to connect the series (setting aside an earlier exploration along the way). I also moved the format and the benefits up the page, where they sat lower on the page it replaced: they are the first thing anyone checks when deciding whether an event is for them.

A texture, and a tool to keep it. I explored ways to control tone in Claude Code and landed on a soft halftone I designed myself, tested across the photography we use. Applying it by hand does not scale, so I built a small internal tool: drop in an image, pick a preset, download the JPG. The first version took about an hour. Try it here.

The agenda page. The agenda is where people decide what to attend, so it had one job: be readable at a glance. I laid it along the timeline of the day, so you get the shape of the whole thing before reading a single description. I coded it as a reusable component too. We run about 60 events a year, so the team swaps in content and the layout handles itself.

Pinned - Web design for sharing local experiences (2026)

Pinned lets people share local experiences on a map of San Francisco. I focused on making it easy to contribute: answer a question, pick a location, and leave a short note. Others can explore the responses by place.^†

† Inspired by Learning to Love You More.

Add your pin ^ㅅ^ Everyone’s welcome.

Clawstin - Designing agentic AI for mobile (2026) Soon

I run OpenClaw, a self-hosted AI agent, on my own server. It can do real work, but I still had to sit at my desk and watch it. Clawstin is the fix.

1. What is it? A remote control for the agent you already run.
2. Who is it for? Founders and engineers who self-host an agent and want to delegate from their phone.
3. Why does it need to exist? Delegation only works if you can walk away. Today’s agents can act, but they can’t tell you when they need you, so you end up babysitting. Clawstin turns that into a clear “your turn” moment.

Jeep 2050 - Future mobility, UX design (2024)

For a Jeep-sponsored project, our team of five explored what a rally vehicle’s interior could look like in 2050. I worked on the interface, focusing on how drivers could read information and use controls at speed on rough terrain. Working with industrial designers from the start taught me how the vehicle’s physical constraints shape the interface, and how visual design, motion, and 3D elements need to work together.

The brief. Bring new life to Jeep’s rally identity without losing what the badge already means. In 2050 it still stands for freedom and going anywhere, though each region says that differently. Five of us worked it, and we set ours in South Africa, where the landscapes and the 4x4 trails made a natural stage for the idea.

What drivers told us. They wanted AI that helped without making the car feel complicated. They wanted the technology, but were not sure it was safe. They liked the idea of rally racing if there were an easier, safer way in. And they wanted navigation that found the places nobody else had been to.

Hands stay on the wheel. On extreme terrain at speed, reaching for a screen is not an option. Voice command became the centre of the concept: you ask, the car answers, and nothing about the road leaves your hands.

Seeing the ground. A holographic scan draws the rocks, ruts and height changes hidden under the car, so the driver knows what is about to happen before it does. Speed, stability and traction sit together in one small cluster, so attention stays on the trail rather than sweeping the dashboard.

Reading the curve. Arrows sketch the safest line through a bend and alerts call out what arrives suddenly. The metric cluster stays exactly where it was, so the view never fills up and the road stays the biggest thing on the screen.

The whole thing, moving. None of it is real until it runs. I built the concept out in motion, frame by frame, so each state answers a situation the driver is actually in: the scan appearing when the ground turns rough, the line drawing itself into a curve, the cluster holding steady while everything else changes around it. The board went up at the show at the end of it.

MiniGloss - A glossary that listens (2025)

Every workplace runs on words outsiders cannot parse. New hires spend weeks decoding them, and nobody wants to stop a meeting to ask what CTR stands for. MiniGloss listens, and names the jargon as it happens. Building it taught me that hierarchy is not a theory: it is font, weight and spacing, argued one small decision at a time. The finished work hung in the Academy's 2025 Spring Show.

The problem. New teammates lose three weeks or more to internal terminology. Meetings stall on definitions rather than decisions. And nobody asks, to save face, so the gap stays open.

Record, and let it listen. You start recording at meeting time. Afterwards the AI pulls the key terms out and shows them twice over: when each came up along the hour, and how much it carried.

A memo while it is fresh. You scan the highlighted words and write against them: takeaways, next steps, what is still open. The words do the remembering.

What testing changed. Initials went from one letter to two, because one could not tell two people apart. Attendee details moved inside each meeting block. And the slide-in panel of definitions came out: testers were watching whether the recording worked, not looking up words mid-playback.

Lingoleap AI - Design system (2025)

Logos, button styles, color choices, and token usage needed clearer rules, and components had become inconsistent. I organized the components and clarified how they should be used, laying the groundwork for a shared system that designers and engineers could build on.

Shades - In-app game design (2024)

Shades was a visual news app for Gen Z, led by Jeff. I joined the team when retention was low and worked on an in-app game feature^† to give people a reason to come back.

I worked through ideas with the team, turning rough sketches into Figma prototypes we could test and refine. From there, I worked closely with the developer and CTO to get it into the app.

By the end, we had a tested game flow and detailed Figma prototypes, with development underway. My work connected how the game played with how readers could find their way back to the news.

† The game: a word association game where players guess a keyword from five clue words, then read the original article to see it in context.

Where to find me.

• LinkedIn
• X
• GitHub