I'm David, a CS major at UW–Madison, with Data Science and Econ along for the ride. I fix my own bike, build my own PCs, and ship my own software. Same itch every time: open it up, find the waste, make it run right.
Four builds, all real, all mine. Grease included at no extra charge.
job no. 001 · feb 2026 - present · thetaviz.com
An options-analysis platform I built because tools for regular investors are either toys or a Bloomberg terminal. Customizable chart indicators are live now; the long game is a platform that shows what investors see in a company and how that's changed: moats, indicators, the story under the price. For any style: value, swing, long, short.
parts used: Next.js · TypeScript · Supabase
job no. 002 · jan 2026 - present · iOS release soon
A nutrition app for UW students. It reads live dining-hall menus (Nutrislice) and turns them into macro-aware meal recommendations. Friendly accountability, never guilt trips. In beta with 40 testers, releasing on iOS soon.
parts used: React Native · Expo · Supabase
job no. 003 · apr 2026 - present · chrome web store
A Chrome extension that drops historical class GPAs (MadGrades) and Rate My Professor ratings straight into UW's registration portal. 100+ installs, all organic. No marketing, just students telling students.
parts used: JavaScript · Chrome APIs · REST
job no. 004 · summer 2026 · UW summer AI lab · 3-person team
Our three-person team built an ML/CV pipeline that turns 3D camera footage into 2D maps of how customers move through a store. Layer LLM workflows on top and you get a small-business consulting platform: detailed graphs that help owners optimize for profit and quality. Onboarding our first test store now.
parts used: Computer vision · Edge ML · LLM workflows
Scroll to reassemble. (I usually have screws left over.)
I can't leave a slow system alone.
A PC that thermal-throttles. A bike that clicks on the downstroke. A checkout line that backs up at 5 pm. A portfolio quietly bleeding fees. Different machines, same problem: something inside is wasting energy. The fix is also the same. Take it apart, find the waste, put it back together better.
That's the actual thread through everything on this page. Computer science is how I open systems up. Economics is how I know what "wasteful" costs. Data is how I prove the fix worked. I'd rather do it myself, even when it takes longer. That's how you learn where the screws go.
operator spec · dmroz
Where I've worked and what got stamped.
Selected for UW–Madison's competitive CDIS/N+1 cohort. Building the retail CV pipeline + LLM consulting platform above, with industry mentors, on a 3-person team.
Interactive visualizations from Census, public, and transit data to analyze community access by ZIP code, with findings presented to Boys & Girls Club leadership.
Ran a STEM nonprofit for four years (~200 volunteer hours), directing 7 instructors teaching Python, JavaScript, and Arduino to 5th–6th graders.
Customer-engagement strategy pitched to Kohl's leadership in a 9-minute final.
Early-talent program participant.
Click a drawer. Everything's roughly where it should be.
The non-billable hours.
Internships, collabs, a stubborn system that needs opening up, or just to talk bikes. Email is fastest.
hours: whenever · reply time: fast · coffee: yes