Josue Montero, Systems Engineer

Managing a logistics company's fleet is harder than it looks. Beyond getting packages from A to B, there's an entire layer of operational complexity underneath: Which driver takes which route? Is that truck due for maintenance? Who approved this work order, and has it been confirmed?

This system was built to answer all of those questions in one place.

The platform centralizes fleet management across a well-defined role hierarchy — administrators, warehouse supervisors, workshop chiefs, mechanics, and drivers each interact with the system differently, seeing only what's relevant to their work. Daily operations flow through a structured order lifecycle: creation, confirmation, assignment, execution, and validation, keeping every stakeholder informed at each step.

The most technically interesting piece was the predictive maintenance engine. Rather than waiting for a vehicle to break down, the system runs periodic analysis across several vehicle metrics — mileage, usage patterns, part history — and automatically triggers alerts when a vehicle is approaching a risk threshold. This shifts maintenance from reactive to preventive, reducing downtime and giving the workshop chief a real picture of fleet health before problems surface.

On the infrastructure side, the application was containerized and deployed using Docker, with Laravel powering the backend, React on the frontend, and PostgreSQL as the data layer.

The result: an entire logistics system built for Correos de Costa Rica — across all branches — that no longer tracks its fleet on paper, and vehicles that get maintained on schedule rather than after the fact.

Some courses push you to think differently. This was one of them.

Racket Mail is a Gmail-like email application built as a university project for a programming paradigms course — the catch being that Racket is a functional language in the Lisp family, far removed from the object-oriented patterns most web apps are built on. No classes, no objects, no frameworks. Just functions, recursion, and a different way of thinking about state.

The entire stack lives in Racket: routing, session handling, and even the HTML and CSS are generated directly from the language itself. Emails are stored as plain text files, one per user, keeping the persistence layer deliberately simple while still supporting a full inbox experience — received mail, sent items, trash, and an all-messages view, with a search bar across all of them.

Users can log in, compose and send emails, reply to threads, and manage their inbox — all through a web interface that you wouldn't guess was rendered by a functional language if nobody told you.

The real challenge wasn't the features. It was rethinking how to structure a stateful, multi-user web application without reaching for the usual tools.

This one started as a visual idea that refused to stay in my head.

Fluid Light Pattern is a React component that renders a glowing blob which follows the cursor — acting as a dynamic light source that reveals animated background patterns hidden beneath the surface. The effect is layered: a canvas of moving patterns sits underneath, and the blob cuts through a top canvas like a spotlight, letting the patterns show through only where the light falls.

The physics behind it are intentionally complex. The blob isn't just chasing the mouse — it has weight, trail, and momentum, built to feel organic rather than mechanical. Getting that to run smoothly required careful attention to performance; canvas animations can degrade quickly, and this one holds up.

It's the kind of component that's hard to describe and easier to just see — which is why it's running live on my own portfolio.

Available as a package via pnpm — and already running live on this very portfolio.

Aniku started the way the best side projects do: a few friends mentioned they'd find it useful, and that was reason enough to build it.

The name comes from anime + niku - Japanese for meat - a nod to One Piece's Luffy and his legendary appetite. The logo is a bitten piece of meat. It made sense at the time, and it still does.

Under the hood, Aniku fetches its data by reverse-engineering a public site's API - decrypting obfuscated URLs to extract video sources and metadata. It's the kind of problem that's equal parts frustrating and satisfying to solve.

The app is organized across five tabs: a home feed with trending titles and genre browsing, a search tab, a weekly schedule broken down by day so you always know what just aired, a Watching section that tracks your progress and lets you pick up exactly where you left off, and a lists tab for organizing titles however you want. A profile tab handles authentication via Firebase, which backs the watch history, lists, and downloaded episodes.

Yes - downloads. Episodes can be saved for offline viewing, with a built-in player for both streaming and local playback.

Built with React Native, designed to feel like a proper streaming app, and made specifically because the alternative was using something that wasn't quite right.