Sustainability and companion planting are some of my big interests right now. I've also been looking into the (relatively) new worlds of computational botany and ecology. I decided to do a programming project centered around this, so I came up with a data warehouse that curates novel companion planting groups. Companion plants are those that provide some mutual benefit to each other when planted together, like the story of the Three Sisters. I'm using Python to iterate through plant data and discover potential mutually beneficial partners for them. This is one of my current projects so this section is likely to change as I progress. Stay tuned for more!
This is a pocket roguelike for mobile devices. I wanted to distill my favorite mechanics from party-based RPGs like Pokémon into a quick survival game. The player manages 4 characters with unique stats, skills, abilities, and some number of item slots. They're thrown against random challenges that the player must navigate to get a high score.
I based all of the characters, events and themes in this game from specific Native American cultures. It was a nice challenge researching concepts from each represented nation and incorporating them into the game. Some cultures (Taíno, Oceti Sakowin, Muisca) are represented in the party members. Other concepts (like the Three Sisters and various endemic animals) found their way into the random events.
I'm working to publish on the Apple and Google app stores, but in the meantime you can download the APK here.
Midway through high school I was inspired by various YouTube videos on AI simulations, so I decided to create my own. I loved history, and I wanted to watch procedural empires rise and fall as directed by a simple neural network. So over the years I've created a series of four programs that I call civilization simulators. They're all written in Java, with the exception of the second (whose final version was done in Python). Each iteration on the project has its own set of unique features, so I'll describe each one.
CivSim1 was the most basic of the simulators. It was also the first neural network I ever wrote. The simulated histories weren't all that complex, but you could watch empires rise and then fall after consuming all the food on their map. Cities would spawn armies, which would go out and either fight each other or found new cities.
CivSim2 was the most popular of the simulators, since it's the first one I shared on my YouTube channel. I scripted a number of conditions that gave this iteration way more life than the first one. Small kingdoms would eat each other to form empires, then those would split up and the cycle would begin anew. The AI was also drastically different, being more inspired by an octopus rather than a traditional neural network. Each AI (emperor) would be given a list of requests from their cities. The emperor's brain would be a set of weights that prioritize these requests based on various factors, since the were only allowed to fulfill so many during each turn. This made for more chaotic simulations, which I thought better resembled real history.
CivSim3 was easily the most complex iteration. I built this one with two modes - the standard auto pilot where you could watch simulations unfold, but also an interactive mode where you took control of one of the initial empires. The goal of the player then was to conquer all the AIs on the map. The game would actually log every decision the player made, and used them to train the next generation of AIs. I wrote a Ruby script that took your play history and tried to emulate your play style with new AIs.
CivSim4 (aka WorldSim) is my latest and prettiest iteration. I had wanted a little less chaos and more interesting histories with this iteration, so I simplified a lot from version 3. The world map displayed less information but there's something so hypnotic about watching those colors move across the screen. And even though it's a little buggy, I still love this version of the project.
I used to work at a computational biology lab as part of the University of Maryland Institute for Advanced Computer Science (UMIACS). Throughout my time there I was able to have my name published in two papers - one about cancer and one about heart disease. For my part in the cancer project, I developed a webapp known as SPAGE that allows users to visualize the paper's data. This data details how different genes interact in cancer patients and their effect on survival rates.
Later on I would head my very own project, developing a tool called FEST (Fast Enhancer Search Tool). The tool would be able to identify functionally-similar genes within a target genome. It would better help us understand mutations and how different species came to be.
This was always one of my sillier projects but I'm proud of it nonetheless. I was hungry in computer networking class one day when I got the idea to write a hashing algorthm that converts files into tasty-looking donuts. At the time I really wanted to learn Rust and WebAssembly, so I wrote the initial version as a Rust CLI tool and then ported it to browsers. You can actually go to this site and convert any file you upload into a donut with sprinkles. Don't be shy, give it a try!
Xochitl and Moonshot are both programming languages that I developed. Xochitl was an experimental passion project from early on in college. I used Node.js to create a compiler and IDE for a wholly unique language, one that almost looks like a logical language but it compiles into SVG images. The goal was to be able to create mathematically precise logos without using heavy UI tools like Inkscape. The syntax of Xochitl was unique in that you could write the lines in any order you'd like, without any effect on the final output. Each line was processed as a node in a graph, so their precise order was unimportant. I wrote the IDE in Electron using the CodeMirror library.
Moonshot was a much more tame idea. I simply wanted a strongly-typed variation of the Lua programming language, something like what TypeScript was for JavaScript. I started by implementing my own Lua parser in C, and then adding on features as I wanted them. I made sure to create new keywords as I went to keep with Lua's overall feel. There are other projects out there that add type checking to Lua, but they also change the syntax drastically for the new features. I was looking for something a little more idiomatic.