Projects

Not the most technically complex thing I've built, but probably the one I paid the most attention to detail on. The old site had two problems: it looked bad, and any content change - a new article, updated info - required a developer. That's not a sustainable setup for a working lawyer. CMS was a requirement from day one. Picked Keystatic because it's local, file-based, and the needs here were genuinely simple: edit content, publish articles. No reason to bolt on a full headless CMS for that.

The actual hardest part was the data migration. The old site's content was raw HTML - inconsistent, messy, not fun. Getting it into a shape Keystatic would accept took scripting and a couple of correction passes. I also did TDD for most of the build, mostly to actually practice it rather than because a static site demanded it. It held up - no surprises, which is exactly what you want from tests on a client project. Mobile-first this time, properly, from the start.

Pair project, tight deadline, and OCaml - which the course technically assumed you already knew going in. I did not. Writing a compiler in a language you're learning while also working through LLVM IR, which sits close enough to assembly that the gap between 'I understand this' and 'I can actually debug this' is significant - that's a specific kind of uncomfortable.

It's also one of the most technically interesting things I've built. You don't really understand data structures until you've had to implement records, tuples, and lists in a language that makes you think about memory layout explicitly. And LLVM IR makes you appreciate every single abstraction sitting above it.

Backend-heavy team project, Azure at the center of everything. The actual functionality is fairly simple an auction platform - but the goal was chaining as many Azure services together as possible: App Service, Cosmos DB, Blob Storage, Redis, Functions, Databricks. Azure is a lot. The docs are dense, the error messages are cryptic, and there was a stretch trying to get Azure Functions deployed for the first time where I was stuck in an error loop I genuinely couldn't get out of. Once it clicked it was fine - but that was a rough few days.

The part that stayed with me most wasn't Azure though. It was the Kubernetes re-deployment at the end to cut vendor lock-in. After weeks of wrestling with cloud-specific configs, Docker and Kubernetes just made complete sense. Came out of this one with a clear opinion on containerization.

The project I've put the most time into so far. Team project, but I owned essentially the entire frontend - somewhere around 85-90% of the UI. Early on I made a call to take on logic that technically belonged on the backend, I volunteered to push myself and for the sake of experiemnt did it, knowing full well it wasn't the right architectural call. Was it? No. Did it push me further than staying in my lane would've? Yes.

The bigger growth was around design. Keeping UI consistent across a lot of pages, deciding what to even expose to users, making those calls over and over. The main thing I'd change: I should've gone mobile-first from the start. Scaling a desktop layout down to mobile is genuinely painful, and I did that the hard way.