← Back to Explore

What I've Built

Projects and the messy paths that got them there

Maersk logistics (solo)

Months out of bootcamp. Solo developer. Enterprise logistics for Maersk. Complex API integrations, German customs systems, XML parsing, airline industry algorithms. Data for 15 distinct offices.

Should have been a team of 5-10. It was me.

The numbers I still remember: cleaned a database from 13.3 million records to 406,000. Analyzed an import integration spanning 1200+ lines. Proposed infrastructure consolidation saving €90/month.

Did on-site user research with five stakeholder roles. Found that one workflow forced 23 manual updates when users couldn't exclude items from bulk actions on 25 shipments. Nobody had asked the users before.

The lesson: institutions often know less about their own systems than you'd expect. Documentation pays off.

Panopticron

Monitoring system for a 6-person agency. Real-time sync from Vercel and GitHub. Self-monitoring. Refine + Next.js + Supabase.

Built it because nobody wants to check four different tabs to see if something broke. Now it's one tab.

panopticron.vercel.app

Uroboro

Knowledge work insights kept getting lost. Good ideas at 11pm. Gone by morning. Tried various note systems. None stuck.

So I built something for 10-second capture. Local AI processing. No cloud dependency. Started with 17 commands and 1558 lines. Refined to 3 essential CLI commands.

Turns out simpler is harder. But it's what I actually use.

uroboro.dev

Examinator

Needed to learn security concepts. ADHD means traditional study methods don't work well for me. Boring = impossible.

Built an offline-first study tool. 186 flashcards, 7 summary docs, local LLM for generating practice questions. Spaced repetition with some gamification that doesn't feel insulting.

github.com/QRY91/examinator

The pattern

Hit a problem. Document what's actually wrong. Build something to fix it. If it works, keep using it. If it helps others, even better.

The spite helps. When institutional knowledge gets punished or lost, documenting it becomes survival strategy.

Everything I build shares DNA: SQLite coordination, CLI-first, local when possible. Not because of ideology. Because it works and I can fix it when it breaks.

Related

Career realignment → AI development journey → Full resume →