Aaron Vanzyl
2025 Resume
I’m Aaron Vanzyl, 22 years old. My main interests are Artificial Intelligence and Unity C#.
Machine Learning
Previous
Next
ML Coursework
Machine learning is a huge interest of mine, especially for creative or artistic content generation. I completed my first machine learning course at 13 and I have now finished a total of 11 courses.
- Machine Learning
- Neural Networks and Deep Learning
- Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization
- Structuring Machine Learning Projects
- Convolutional Neural Networks
- Sequence Models
- How Google does Machine Learning
- Launching into Machine Learning
- TensorFlow On Google Cloud
- Feature Engineering
- Machine Learning in the Enterprise
Markov Chain Music
A music generator using Markov Chains. The generator uses variable order Markov chains trained on any input MIDI file (including multiple instrument tracks) to create an endless amount of new music.
The attached video describes the project. If you open the video in youtube, it has extra links to recordings of just generated music in the description.
Previous
Next
Pokemon Image Generator
A Pokemon image generator made with a variational autoenconder. The model learns to compress images to a small vector and then convert that vector back into the same image. Once this is learned, inserting random vectors in the middle step lets the model decode entirely new images.
For the encoding part, I used style loss rather than regular pixel loss. This means that instead of maximizing pixel similarity, the model aims to maximize similarity in the activations from the vgg16 image classification model. These activations capture broader patterns or style in the image, and results in the images produced being more consistently similar to the source in style (but not necessarily content).
Previous
Next
Learning Agents in Prisoner's Dilemma
A simulator for a large population of evolving agents playing prisoner’s dilemma. Over time, the agents modify their strategies in response to the rest of the population.
I investigated how changing the agents’ initial strategies and the game reward function affects the growth of the population.
- Does the population always collapse into total cooperation or total defection?
- How can we increase the probability that the agents will settle on cooperation?
Unity - C#
Previous
Next
Terrain Generator
A generator that creates large-scale terrain using height maps and simulated biomes. Firstly the height is generated using several layers of Perlin noise. Then, the generator uses a combination of temperature and humidity simulation to assign 8 different biomes across the map.
Includes rivers, lakes, cities, and a lot of trees.
Previous
Next
Multiplayer Spaceship Game
A multiplayer game where players build and fly around physics-based spaceships.
Features
– Online multiplayer.
– Fully physics based. Ships can collide and bounce!
– Turrets and lasers.
Live web version (If you want to connect twice on a single computer, use a different browser or incognito window)
Previous
Next
Walrus Adventure
A procedurally generated 3D rogue-like where you play as a laser-shooting walrus with a tophat.
Key features:
– 4 different level types, each with different terrain, enemies, and bosses.
– Enemies have different AI, stats, and actions.
– XP and leveling system that changes player’s stats and projectiles.
– Fire that destroys certain blocks and spreads.
– First-person and top-down views.
– Runs on both desktop and Android using tilt controls.
Art & Others
Previous
Next
Maze Art Collection
A procedurally generated maze-inspired NFT collection. Each image is created by leaving a trail of colors while navigating through a randomly generated maze. The collection contains single-color, rainbow, and monochrome maze variants.
Previous
Next
16 bit CPU and GPU
A 16 bit computer made in Logisim. Includes a CPU, ALU, GPU and Register bank – all constructed from basic logic components. I also designed an assembly language and built a compiler that converts assembly code into binary that can be run on the computer.
ABOUT ME
My name is Aaron Vanzyl, I’m 22 years old, and I have been programming for 10 years. I am currently studying computer science at Monash University. I’ve lived in Australia, Bangkok, and the United States, and spent a year traveling in 2016-17 (which is where most of the cool photos are from).
I have experience in:
- Python
- Java
- C#
- Unity
- Machine Learning
Machine learning is a big interest of mine, and I see a lot of interesting application for it in generative models, such as art or music, and for artificial intelligence in games as challenging opponents or more interesting NPCs.
I am interested in AI/programming jobs or internships in Melbourne. If you have any opportunities or would like to talk, please email me at aaron@vanzyl.org
In my spare time, I play video games and guitar, and do a lot of skiing. My favorite games are Skyrim, Divinity OS 2 and Sekiro.
