Marco Ferreira — explore the machine

Setting up the desk

What's inside

The Core: Who I am, how I got here, and what I do away from the keyboard.
Accelerated Ideas: The AI systems I build at work, and the models I run at home.
Active Memory: What I'm working on right now, and the skills I have.
Project Archives: My biggest projects, what I'm building now, and everything else by category.
Connected Systems: Software built to scale, and the systems I've built to learn how.
Connect: How to reach me, plus my résumé.

Projects

Draftly: A Google Docs-style editor where several people can write in the same document at once, which I built with my friend Richard Baah.
OptiSQL: An in-memory query execution engine I built with my friend Richard Baah.
GooseDB: A full database system I'm building from the ground up in C++23, for storing and querying media.
myshell: A Unix shell written in C, with pipes, redirection, wildcards and a batch mode.
mymalloc: My own malloc and free in C, with error detection, leak reports and a benchmark against the real thing.
Cache simulator: A CPU cache simulator in C that replays memory traces and counts hits, misses and evictions.
Recursive word count: A C program that walks files and folders and lists every word by how often it appears.
Pratt-parsing expression evaluator: A C++23 calculator REPL that parses expressions with Pratt parsing and remembers variables.
IMP interpreter in OCaml: An interpreter for a small imperative language, with loops, conditionals and functions.
Face and digit classification: Three classifiers for handwritten digits and faces, two of them built from scratch, compared on accuracy and speed.
A* search through mazes: Repeated Forward, Backward and Adaptive A* finding paths through mazes the agent can't see in advance.
JBMDL assistant, first version: The first version of the AI assistant I built for Joint Base McGuire-Dix-Lakehurst: ask by text or voice, get cited answers.
Python executor MCP server: An MCP server that lets AI agents run Python code, with the dangerous parts blocked.
House price regression: Predicting house prices from their size with linear regression and a decision tree.
Distributed training system: A fault-tolerant distributed training system in Go, with a replicated parameter server.
This portfolio: The site you're on: a 3D desk with my PC, where every part opens a section on the monitor.
Solar System Explorer: Fly a spaceship through a solar system with real orbits, in the browser.
RU Burgers: A restaurant ordering app, built twice: once for the desktop with JavaFX and once for Android.