Hey, I'm Alex

I built this because most game recommenders stop at "pick a game you like" and then hand you a mystery list. I wanted something that lets people say what actually matters: Persona 5 for the jazz fusion OST and modern Tokyo setting, not just because it is an RPG. The goal is to help niche games surface for the right reasons, and to show users why a match made sense instead of hiding the recommendation behind a black box. It is free, open source, and very much a passion project; the implementation details live in the repo for anyone who wants the deeper data-flow tour.

An explainable wine recommender developed with the Superlinked team during early access to their SIE engine. Instead of only returning similar bottles, it shows which attributes shaped the match, like fizz, cherry notes, body, or acidity. The demo uses document processing, vector embeddings, and small-model inference to make wine discovery feel more transparent.

A stochastic pandemic simulator that models disease spread across city coordinates using Markov-chain mobility and SIR dynamics. Python computes the population state transitions, then Godot renders the spread on a 3D globe so people can see the model instead of just reading equations. It earned 2nd place at BLOOM Hackathon and later received grant support for neural-network forecasting work.

A hackathon project built to help people make sense of retirement planning questions like "Can I retire in Asia?" or "How much should I start saving?" The goal was not to automate away accountants, but to make the first pass through retirement variables faster, clearer, and less intimidating. We built a modular RAG system in about five hours with no WiFi, with finance guidance from Richard Vo at AMEX and teammates Eric Taylor and Faadil Shaik.

A high-throughput crawler built around data movement more than page scraping. 300 extract workers download and parse pages while 300 database workers batch writes into SQLite, with queue buffers keeping the system fed without falling over. It was a fun way to think through throughput, batching, worker coordination, and what happens when your pipeline moves faster than your database wants it to.

An open-source ML security tool for finding suspicious or poisoned content in training datasets before that data quietly becomes model behavior. The system uses a multi-agent analysis pipeline to inspect dataset content, generate hypotheses, and surface examples worth investigating. I like this project because it sits right at the point where data quality, automation, and model safety stop being abstract and start being operational.

A gamified Canvas LMS tracker that pulls assignments into a game loop, then scores urgency using completion patterns and difficulty signals. The fun part was not the coins; it was turning school workflow data into something students could act on without needing another dashboard yelling at them.

A hackathon-built 3D Godot experiment in procedural space generation. We used swirled Perlin noise for planet placement so the world can keep expanding with an infinite-feeling galaxy structure, wave function collapse for city placement, and procedural systems for enemies and asteroids. It was not continued after the event, but I keep it here because it shows my interest in systems that evolve from local rules.

A one-day game jam project we made and demoed at GDC Festival of Gaming. It was my first time at GDC, and showing a tiny game to a room full of people who love making things was genuinely special. Built with Keriya Son on 3D, Angie Peczon on art, Eric Taylor on shaders, and Ilce Perez on music. It was not continued after the jam, but the artifact captures the speed and joy of the event.

A small vertical climber built over a weekend for a UC Merced game jam in 2022. It was my first project and was not continued after the jam, but it is still here as a reminder that shipping small, complete things is its own skill.

A small Obsidian community plugin for cleaning up writing without leaving the note. It helps catch repeated misspellings, apply quick corrections, and keep a personal dictionary for words Obsidian should stop fighting you on. It is less glamorous than a big data system, but it has the same shape I like: take a messy stream of text and make it easier for people to work with.

A dietary search app that filters restaurant menu items by nutrition and constraint data. It is a smaller project, but it still has the shape I keep coming back to: pull in messy real-world information, normalize it enough to search, and give people a cleaner way to make a decision.

A small visualization for spiral growth using queued directions. It is a compact teaching project, but I keep it around because visualizing state changes is often the fastest way to understand a system.