← Back to work

Internal software · 2026

Aspen OS

The project operating system the QuickRN team actually runs on — tasks, kanban, documentation, and collaboration in one place. Built solo, in daily use at a nursing-education company.

daily use at QuickRN
in production
versioned SQL migrations
42
Playwright end-to-end specs
32

Architecture

Select a stage to explore it.

01 · Next.js app

Server-first React with the App Router: shadcn/ui components over Tailwind, client interactivity only where the kanban and editors need it.

TypeScript strictSQL migrations in version controlbuilt for and with the team using it

Visuals

Aspen OS application showing the project dashboard
Live application — internal project operating system in daily use at QuickRN

01 · Problem

QuickRN is a nursing-education company, and its project coordination lived across spreadsheets, chats, and memory — the default state of most small teams. The team needed one place for tasks, boards, and documentation, simple enough that non-technical members would actually use it.

02 · Solution

Aspen OS: an internal project operating system built on Next.js and Supabase, covering task management, kanban workflows, project tracking, documentation, and collaboration — designed with the people who use it daily.

03 · Engineering decisions

Server Components by default with client code only where interaction demands it — the same discipline as this portfolio. A relational Postgres schema with versioned migrations instead of a schemaless store, because project data has real relationships and real integrity constraints. And 'simplest tool that the team enjoys' as the explicit product bar, which cut more features than it added.

04 · Measuring before optimizing

When the app felt slow, I profiled production with Playwright tracing, V8 code coverage, the bundle analyzer, and the React profiler — and then audited an earlier performance write-up against those measurements. Several claims did not survive: a duplicate react-dom said to cost 1,200 ms measured about 83 ms of parse time, and a suggestion to swap server-verified auth for a cheaper session check was reversed, because the security boundary is worth its ~130 ms. JavaScript parse and execute, not React re-renders, was the real bottleneck.

05 · Results

The platform is the team's daily hub at QuickRN — the strongest kind of result a portfolio project can have: real users who chose to keep using it. Built end-to-end solo, from requirements gathering with non-technical stakeholders to deployment and maintenance.

06 · Lessons learned

Products for real teams are shaped by adoption, not features: every design decision had to survive contact with people who don't care how it's built. Maintenance — the part after shipping — is where most of the engineering lessons lived.