I build for the web, then I make it survive production. My day-to-day is TypeScript end to end — React and Next.js on the front, Prisma and Postgres behind it, deployed to the edge. Everything I ship is written to be tested, typed, and readable six months later.

What I'm chasing right now is the layer under the app: system engineering, Docker, CI/CD, and Linux internals — because I don't want to just ship features, I want to understand exactly how they run.

A local-first SEO inspector for Chrome, Edge, and Firefox. No account, no server, no page URL ever leaves your machine.

• Audits titles, meta descriptions, canonicals, indexability, Open Graph / Twitter cards, heading order, links, and word count — with a live SERP preview.
• Site mode discovers pages from sitemap.xml (internal-link crawl as fallback), then scans up to 1000 in the background with a progress bar that survives a service-worker restart.
• Polite by design: robots.txt and Crawl-delay are respected, 4 parallel requests, hard page cap. Reports export to CSV/JSON.
• One self-contained DOM extractor is shared between the live tab and the crawler, so a page is never audited twice by two different sets of rules.

Every US state and territory flag as code — because no Unicode/emoji equivalent exists for them.

• Ships as raw SVG, React components, JS strings, and CSS classes from one generated source of truth.
• Preserves each flag's native aspect ratio — state flags are not uniform (Ohio is a swallowtail pennant, Rhode Island is nearly square). 16 distinct ratios, handled correctly by every entry point.
• A build pipeline that goes from curated source SVGs → optimized assets → generated React/CSS/string bundles → a validated exports map.
• Ships behind a real quality gate: unit tests, publint, Are The Types Wrong, and an install-the-tarball test — all wired into prepublishOnly and CI.

open-editor — a local-first video editor. No account, no upload, no render farm. Planning the architecture: one Canvas 2D + WebCodecs renderer shared by preview and export, flat-JSON projects instead of a compiler, and FFmpeg scoped to decode only so an LGPL build never poisons the license.

• 🔧 Learning system engineering properly — processes, signals, networking, and what actually happens between npm run dev and a running container.
• 🐳 Getting hands-on with Docker: multi-stage builds, slim images, compose, and why your image is 1.2 GB.
• ⚙️ Building CI/CD pipelines that actually gate: tests, lint, typecheck, build, publish.
• 🐚 Deepening Linux fundamentals — because debugging at 3 a.m. is easier when you know what the kernel is doing.
• 📐 Reading and planning around open-editor, a local-first video editor built on WebCodecs.

"Ship it local-first, test the parts that scare you, and never make the user wait on a server they didn't ask for."