I am a game designer, computer scientist, and game developer, with a fondness for both the realistic and abstract visuals that can be produced by computers. I recently completed my Master's thesis in Computer Science at Rensselaer Polytechnic Institute as part of the co-terminal program, a 4+1 bachelors and masters program. My undergraduate degree is in Computer Science with a dual major in Games and Simulation Arts and Sciences. My research focused on physics-based animation of a humanoid character, as described in more detail here. Other areas of interest include non-photorealistic rendering, realistic lighting techniques, landscape and city generation, participating media, and hardware acceleration of graphics techniques.
Game Design
My game design interests are varied, but I tend to create games with short storylines that incorporate an interesting visual effect, usually with puzzle-based or competitive games with very little story. I enjoy incorporating non-photorealistic techniques into games, such as the watercolor effect used in Mar. In past situations, given a game project to be completed, I have found a way to make the project something I enjoy and will be proud of, whether the genre, style, or type of game is something I normally play or not. One of my ambitions is to create an MMORPG, though I also aspire to work on a competitive game and short story based titles.
Outside of Work
Outside of work I am an avid martial artist and have practiced Tae Kwon Do, Eskrima, and most recently Capeoira. Additionally I amuse myself with origami, playing music on flute and piano, and fiddling with various code projects. I am also a gamer, playing League of Legends consistently, though I have in the past played numerous MMORPGs such as FlyFF, Cabal, Rappelz, Megaten, Lord of the Rings Online, and Guild Wars 2. I have also played a fair share of Team Fortress 2 and Terraria, and made my way through a number of single player games, my favorites of which include Spyro the Dragon, Jak and Daxter, Mirror's Edge, Antichamber, Limbo, Borderlands, Thomas Was Alone, and Dishonored. The list of games I like and have played is constantly growing and changing.
If AI has a future (a big if), it will have to be economically viable. An industry can’t spend 1,700% more on Nvidia chips than it earns indefinitely – not even with Nvidia being a principle investor in its largest customers:
A company that pays $0.36-$1/query for electricity and (scarce, fresh) water can’t indefinitely give those queries away by the millions to people who are expected to revise those queries dozens of times before eliciting the perfect botshit rendition of “instructions for removing a grilled cheese sandwich from a VCR in the style of the King James Bible”:
Eventually, the industry will have to uncover some mix of applications that will cover its operating costs, if only to keep the lights on in the face of investor disillusionment (this isn’t optional – investor disillusionment is an inevitable part of every bubble).
Now, there are lots of low-stakes applications for AI that can run just fine on the current AI technology, despite its many – and seemingly inescapable errors (“hallucinations”). People who use AI to generate illustrations of their D&D characters engaged in epic adventures from their previous gaming session don’t care about the odd extra finger. If the chatbot powering a tourist’s automatic text-to-translation-to-speech phone tool gets a few words wrong, it’s still much better than the alternative of speaking slowly and loudly in your own language while making emphatic hand-gestures.
There are lots of these applications, and many of the people who benefit from them would doubtless pay something for them. The problem – from an AI company’s perspective – is that these aren’t just low-stakes, they’re also low-value. Their users would pay something for them, but not very much.
For AI to keep its servers on through the coming trough of disillusionment, it will have to locate high-value applications, too. Economically speaking, the function of low-value applications is to soak up excess capacity and produce value at the margins after the high-value applications pay the bills. Low-value applications are a side-dish, like the coach seats on an airplane whose total operating expenses are paid by the business class passengers up front. Without the principle income from high-value applications, the servers shut down, and the low-value applications disappear:
Now, there are lots of high-value applications the AI industry has identified for its products. Broadly speaking, these high-value applications share the same problem: they are all high-stakes, which means they are very sensitive to errors. Mistakes made by apps that produce code, drive cars, or identify cancerous masses on chest X-rays are extremely consequential.
Some businesses may be insensitive to those consequences. Air Canada replaced its human customer service staff with chatbots that just lied to passengers, stealing hundreds of dollars from them in the process. But the process for getting your money back after you are defrauded by Air Canada’s chatbot is so onerous that only one passenger has bothered to go through it, spending ten weeks exhausting all of Air Canada’s internal review mechanisms before fighting his case for weeks more at the regulator:
There’s never just one ant. If this guy was defrauded by an AC chatbot, so were hundreds or thousands of other fliers. Air Canada doesn’t have to pay them back. Air Canada is tacitly asserting that, as the country’s flagship carrier and near-monopolist, it is too big to fail and too big to jail, which means it’s too big to care.
Air Canada shows that for some business customers, AI doesn’t need to be able to do a worker’s job in order to be a smart purchase: a chatbot can replace a worker, fail to their worker’s job, and still save the company money on balance.
I often heard “you’re a programmer, AI is coming for your job”. I never agreed with it and this article greatly presents why.
I mean - yeah, technically if you don’t care about stuff like security, reliability, bugs and such, then you can use AI to generate your code. That’s assuming you’re coding something, that was already done a lot of times, and even then you need someone go review it… I wonder who that might be…
So AI - or to be exact: LLM solutions, as this is what this article focuses on - is not going to take my job. Under misguided management it can make my job a living hell of only code reviews and bug fixing AI generated code, but programmers will be still necessary.
And with inevitable decline in code quality of AI solutions, software development may actually take up LONGER to create if you want it done mostly by AI. Or at very least - it will become much more expansive to maintain. Because fixing and maintaining AI generated code will be harder and it will much more prone to bugs.
As an automation expert I couldn’t agree more. This is even true for coding specific LLMs like Copilot. I generally avoid them. My colleagues don’t seem to understand why as of yet. I might send them this essay.
an interesting take on this from matt braly, creator of amphibia
Imagine being a business that has managed to get THE MAXIMUM NUMBER OF CUSTOMERS IT IS POSSIBLE TO HAVE, and you’ve managed to make it so this is a bad thing.
You did it. That’s late stage capitalism in one sentence.
society is spiraling and culture is a wasteland. i know this because i looked and most people prefer things that are fun and easy, making fun and easy things extremely popular. this is the first time that’s ever happened, historically.
in order to get people to enjoy things that are challenging, i will find the people who are enjoying fun and easy things, and i will tell them how cringe and stupid they are
where relevant i will pirate fun and easy things, the better to explain why they are cringe and stupid
if i accidentally find myself enjoying those things, i will find ways to make them difficult through essay writing
today i am thinking about dracula daily. i am thinking about how so many people signed up not because they wanted to assign themselves homework, but because people told them it was going to be fun. or they saw other people who had signed up having fun, and they wanted to get in on the fun. i am thinking about people who started dracula daily shitposting about paprika and shipping jonathan, and ended it writing and reblogging essays and book reports about an 1897 novel–which was still for fun. and i am thinking about people who still think it doesn’t count, that all those people weren’t really reading classic literature, and the evidence is that they were having fun.
i am thinking about how when i had my gallbladder removed, i was supposed to give up fatty foods, but mostly i didn’t. this is attributed to my lack of willpower. i’ve had horrible heartburn since i was in elementary school, but everyone understood why i didn’t want to give up tomatoes. tomatoes are good for you. i fucked up my knee doing squats and spent months fucking it back up because i tried to get on the treadmill again as soon as i could get away with it. this is not attributed to my lack of willpower.
it’s just so weird. even aside from the ahistorical nature of it, the weird idea that no one read romance novels and watched action movies and sitcoms before tiktok made them stupid. you should watch better movies, read better books, because you’re having too much fun. watching movies and reading books should be unpleasant. like telling people they should eat their vegetables because dinner is supposed to taste bad.
an experimental queer novella won a bsfa award, a nebula, and a hugo, but nothing drove sales as hard as a vashwood shipper aggressively and enthusiastically telling people they should read it with emphasis on the idea that it wasn’t a big commitment.
when i watch a french film about a succubus or a chadian film about justice, i’m not doing it because movies are supposed to be difficult to watch. i’m doing it because they’re good movies. why would i tell someone that the movies they watch are stupid, and that they might be too stupid for the movies i watch, if i actually wanted them to watch good movies? when i read nonfiction about grocery store supply chains or nixon’s presidency i’m not doing it because i want to better myself. i’m doing it because i’m enjoying myself. i am glued to my couch and i am reading the whole thing cover to cover without stopping to take notes because that’s not how i enjoy books. i am buying a song on bandcamp that only 13 other people have bought because i like it, not because popular music is Too Listenable.
it’s like. if i tell someone how many books i read in a year, there is one kind of person who will be very impressed because they think reading literally anything is for intellectuals, and there is another kind of person who will immediately assume that i am just reading romance novels and be deeply unimpressed. and they are both assigning values to how i spend my time based on whether or not they think i was ‘just’ enjoying myself.
at the end of the 19th century dime novels were massively popular in the united states. they were also mostly godawful. i’ve read them. they sucked. anthony comstock said children were having their minds ruined by reading trash instead of good books, and through various campaigns spearheaded by his new york society for the suppression of vice he is said to have destroyed 15 tons of books. the comstock laws from 1873 criminalizing the use of the postal service to mail obscenity, sex toys, letters with sexual content, or any information regarding those items is still on the books in the united states. the ban on mailing contraceptives and abortifacents was only recently struck down.
Again, these stories breed vulgarity, profanity, loose ideas of life, impurity of thought and deed. They render the imagination unclean, destroy domestic peace, desolate homes, cheapen woman’s virtue, and make foulmouthed bullies, cheats, vagabonds, thieves, desperadoes, and libertines. They disparage honest toil, and make real life a drudge and burden. What young man will serve an apprenticeship, working early and late, if his mind is filled with the idea that sudden wealth may be acquired by following the hero of the story? In real life, to begin at the foot of the ladder and work up, step by step, is the rule; but in these stories, inexperienced youth, with no moral character, take the foremost positions, and by trick and device, knife and revolver, bribery and corruption, carry everything before them, lifting themselves in a few short weeks to positions of ease and affluence.
I vote we stop calling it inflation at all. Seize the language. It’s price gouging, not inflation. Inflation is a nebulous concept that invokes feeling of being too complex for the layman, a struggle as old as economy itself against a beast no one has ever truly slain.
Price gouging is the truth of it. And it makes it very clear who is to blame, and what must be done to end it.
Can confirm this works wonders. Australia is in a cost of living crisis rn and the two major supermarkets are a big part of it, as they pretty much have a duopoly on not just the grocery shopping market, but a bunch of others considered to be essential (things like fuel). They are trying to blame their price rises on inflation, but the media recently started reporting it as price gouging (which it is), and it got the average person pretty worked up, better than blaming inflation did.
It should be illegal to require that any device or software connect to the internet just to run. I shouldn’t need to log in with microsoft to open any of their programs on my local computer. All games should be playable without access to an online server. All media you pay for should be downloadable to local disk as a raw file and if they don’t like that because they know you’ll share it and upload it, tough shit. They took your money already, they’ll live.
The goal was to produce a hand-drawn effect on rendered scenes in real time, comparing combinations of a cel-shaded effect, outline drawing through two different techniques, and a modified version of the hatching technique from Real-Time Hatching. The effects were produced using Cg shaders, with the OGRE 3D engine handling rendering, scene management, mesh and texture loading, and passing of values to the shaders.
We created two algorithms for producing a cel-shaded appearance, each creating step functions to limit the number of shades. The first method applied the step function to the RGB channels, while the second converted the colors to HSL space before applying the function to the saturation and lightness channels. Outlines were produced first through the use of a Sobel filter, applied as a post-processing effect, and secondly by re-rendering the geometry scaled slightly larger and culling front-faces. Hatching was produced with the method outlined in the paper, blending between toroidal mip-mapped textures with different auto-generated patterns of hatch marks to shade the model.
Implementation based on Parallel Controllable Texture Synthesis (Hoppe et al.)
The goal was to synthesize deterministic patches of texture from a small example patch. Our main source was Parallel Controllable Texture Synthesis (Lefebvre and Hoppe), though we deviated from their method due to time constraints (2 week long project). Our final results are still wanting, so we plan to continue work when schedules allow for it.
The general idea of our method is to replicate the patch to produce a larger image, upsampling the patch by making a texture where each pixel corresponds to a coordinate in the exemplar. These coordinates are then jittered, and for each pixel in the generated image, the neighborhood is compared to the exemplar's neighborhood and corrected so as to produce similar patterns to those found in the exemplar. Our method still falls short on producing results that are consistently similar to the exemplar, often producing noticeably anomalous regions. We also do not have any compensation for textures with regular, structured patterns, such as bricks, where warping is noticeable.
Deferred rendering pipeline with basic game engine components
This was yet another project I made with my friend Matt for a class. The idea was to try and get a photon mapping implementation as outlined in Hardware Accelerated Global Illumination by Image Space Photon Mapping (McGuire and Luebke). The deferred rendering component was primarily my doing, with Matt creating the texture loading and error libraries.
At the moment, this is my goto code when I need something rendered. Features include texture loading, obj model loading, direct and deferred rendering, a ray tracing component, shader loading and compilation, and a rudimentary scene graph. There are also libraries for vector math, matrix math, and quaternion calculations, for transforming vertices. The photon mapping implementation is incomplete and as I use this code to experiment, not all of it us functional. The initial push for creating this was about a week, where I wrote over 6000 lines of C++ which Matt helped refine and restructure. Further changes then took place over several days to produce the final product.
This project was done in CUDA C with a parallel construction. The simulation follows three rules as outlined by Craig Reynolds: separation, cohesion, and alignment as per the description of the boid simulation in Flocks, Herds, and Schools: A Distributed Behavioral Model by Craig Reynolds in the proceedings of SIGGRAPH 1987. Visualization of the boids was done using OpenGL, with isosceles triangles representing each boid. Triangles were generated using a geometry shader written by my project partner, Matt McMullan. We had mixed results with this, where sometimes the boids would function normally and sometimes would have no data due to an issue with our usage of CUDA that we never tracked down.
Implementation based on Interactive watercolor rendering with temporal coherence and abstraction (Bousseau et. al)
This implementation was created in Unity3D as part of a game called Mar. After working on the cross-hatching rendering technique I was interested in working with other non-photorealistic rendering techniques, specifically watercolor style. I had in mind a game that would look like a geometric, abstract painting, and after proposing the plan to Matt, we began development, with him focusing mainly on the game mechanic implementations while I found several papers on various watercolor rendering techniques and began testing them out.
A piece of paper on a covered tabletop folds itself into an origami crane and flies away. This was done as stop motion using image matting, compositing, and inpainting. A combination of code written using OpenCV and GIMP plugins were used to produce the final result. I produced the OpenCV code and used several existing GIMP plugins. GIMP was mostly used for touch-ups and for fixing and replacing failure cases for my inpainting code (of which there were many) and to help cover failure cases of matting and compositing. This project focused more heavily on the final resulting video than on the success or failure of the techniques I used.
Setup, data collection (shooting footage), coding, editing, and final video creation were performed over the course of 3 weeks as an individual project. I received consultation on my setup and photography from Mary Gordon, who also loaned me equipment (and kindly helped me in setup and tear-down).
This project was inspired by characters such as Toph from Avatar: the Last Airbender and DareDevil from Marvel Comics who "see" the world through sound or other vibrations. This effect is also seen in recent Batman media (DC Comics), specifically the Arkham series of video games and the film The Dark Knight, where a sonar style system gives Batman a visual of the city. These vibrations are often visualized in media as ripples spreading through the environment which outline objects, showing a stylized view to indicate this unique style of "sight."
For this implementation, a Fast Fourier Transform (FFT) analysis of a sound file was used to generate the ripples, giving a sort of music visualizer. The implementation was made using OpenFrameworks and GLSL. Several test songs were used, including Settle Down by Kimbra which was used for the video below. Due to the computational intensity of both screen recording and producing the output, the video is somewhat choppy.
Animating a character for video games and films is difficult and time consuming, requiring hours of artist labor to produce each animation. These animations are set and inflexible, requiring changes to the animation or sometimes fully new animations to suit new characters or situations for natural looking movements. Jumping is one such animation, where the size, mass, strength, and environment affect the movement of the character. Traditionally these animations are produced by manually posing the character for certain key frames and interpolating between the frames to produce a smooth animation. The more detailed or lengthy an animation, the more work required to specify it. Physics-based simulation for animation production can reduce this work, creating animations for a variety of situations based on constants set for the character and environment. These animations can then be easily recreated or adjusted for different environments by changing the constants set for generation.
This thesis work presents a simulation-based method of control for a character, focusing on the lower body, to produce jumping animations for a variety of situations and body parameters. Two methods of simulation are described, one using a torque calculation and the other using an energy calculation to determine poses for the character. My simulation takes as input a mesh representing the character, a tree of joints describing the skeleton, a set of muscles, mass assigned to each limb of the body, and a description of the desired path through desired timings, gravity, and desired displacement. An inverse kinematic solver is used to aid in posing the character.
Contributions of this thesis include an implementation of a simulation to produce jump animations in \unity{}, a description of character poses based on torque as well as another based on energy, a sampling-based method for choosing a target position, and visualization of the produced animations in several ways to aid in debugging, analysis and presentation.
This project started off severala times as a personal project and was born out of the mess I created when making short webgames. More recently, my goal became putting live examples of my work on this website for easier perusal. I am working to mimic the straightforwardness of OpenFrameworks and Unity scripting so that this can be used for rapid prototyping. There are other WebGL engines and libraries that I could use but part of the reason for this project is for personal satisfaction and learning.
The test page has an error log giving compile and link status of shaders, some debugging information about different objects, a constantly updating feed of the vector bases for the camera (view->world matrix component vectors) and cube model (model->world matrix component vectors for the cube). There's a known issue that I've been trying to pinpoint in which the cube flattens to a hexagon whose plane is orthogonal to the axis of rotation when rotated about certain axes. I've sideboarded the issue as I thought that it was a vector normalization issue or a mistake in my quaternion rotation math, but have yet to create a change in the behavior and have thus moved on temporarily to add more required features so that I may start porting my work and putting it online.
Work on this project is ongoing and I will continue adding features as the need arises.
A game about a lovable stuffed tiger who wishes to become the only one of his kind so he can brag to the other stuffed animals. Developed as a school project in 2 weeks with a group of 5 in pygame. Our team consisted of three programmers and two artists. My main responsibilities were the rudimentary AI for the enemies, the IO and event handling, and the classes used for the projectiles, enemies, and player character, including movement, spawning and interaction. I also collaborated with another of the programmers to handle collisions and implement our custom class for dirty sprite animations, which used the built in drawing functions but replaced pygame's sprite animations. As a result of our optimizations, we had a framerate, as indicated by a counter, of over 1000fps with around 15 moving characters on the screen. We introduced a frame limiter as a result.
The player is able to attack by throwing buttons they collect from the bodies of their vanquished, plush foes. Other attacks include a simple punch and a "ground pound" attack, in which the player character jumps into the air, gracefully executes a flip, and punches into the ground, damaging everything around them. When damaged, they can replenish their health by collecting stuffing, again from the bodies of their vanquished foes.
Samples of work developed using the Panda3D engine. Although I have used the engine and have a moderate degree of familiarity, I wouldn't use it for production, as it is more of a teaching and learning tool. It was important in my skill development however, showing a distinctly more programmatic approach to composing scenes and dealing with game and engine logic than is found in Unity, which has a more graphically based "drag and drop" type of workflow. There is one actual game demonstrated here, while the other examples were part of an individual assignment to produce a simple vehicle (in my case, a penguin with an attached rocket) with a light source that could be moved through the environment to avoid or collide with obstacles (an oversized mushroom and a boulder) in the level.
The individual assignment required each person to produce their own art as well as their own code. I produced the models of the penguin, rocket, mushroom, and boulder, as well as the textures for each, and created a simple environment where a random configuration of mushrooms and boulders were spawned, and the penguin would then be able to navigate. I produced a particle effect for the rocket as well as placed a light for the flames coming from the back.
The group assignment was a game called Alzarian's Artifact, and was based off of the ideas "raise" and "elude" (randomly assigned). The idea is that the player has explored some ruins, and discovered an artifact which allows them to control the stone around them, only to find that there are monsters stalking through the halls. They then elude the monster by raising walls out of the stone beneath their feet, escaping while the monster is stuck, or hiding behind the walls, exploiting the monsters' poor eyesight. The player is able to raise a small wall (a limited number per level) and place small lights to better see the dark rooms while they move to the marked doorways leading to the next level and eventually out. My responsibilities were mainly lighting, as well as a particle effect for lamps, environment modelling,
An unnamed game developed with the Unity3D engine about a robotic dog. It was developed as a class project, and the assigned "theme" for the game was hamartia. Our concept was a robotic dog produced by the military who is loyal and obedient to a fault, progressing through the levels of the game and slowly damaging itself. Eventually, though the dog completes its tasks, it is far too damaged to be useful to the project and the scientists running the project order him disassembled in favor of building a more able prototype.
My responsibilities in this project involved work on customized character and camera controllers, as well as prefab creation, asset imports, level design for one of the levels, asset server management and troubleshooting, general Unity troubleshooting for the whole team, terrain creation, applying collision solids to imported models, animation import and association (with scripts and models) and finally particle effects for a flame effect used by the dog for a flamethrower (used for clearing brush obstacles from the player's path).
Supernova
Project duration: Ongoing
Team size: 2 (potentially going to increase)
A game in the early stages that I'm developing with my friend Rosa Tung. We're currently working in Unity to develop this somewhat abstract game, in which you play in a zero gravity multiplayer environment as a sphere, with the ability to absorb small particles of light. These light particles can be shot at other spheres or used as an attack. Absorb too many light particles and you'll explode, but too few and you implode.
Some of the effects we're working on are a "fat cell" type of look for the spheres, giving them a membranous feel. We also aim to experiment with different control schemes to produce a control scheme which immerses the player and enables them to navigate and attack easily.
Subterrane is an underground horror-exploration game. You play as an athletic caver who has slipped unexpectedly into a deep hole. Your journey to climb out takes you deeper into the earth where you find some of the creatures who live in the caves. This game was developed in Unity3D with a team of 5.
The material applied to the crystals and mushrooms was provided courtesy of Anton Hand.
Particle effects and fog (fog at beginning/end of levels, depth fog)
Mar is an abstract physics game about shepherding the mar balls through their dangerous, puzzling environments, inspired by Lemmings and World of Goo. The mar proceed according to a few simple rules:
Start moving right
If they hit an obstacle or wall, reverse direction
The mar can be altered to help their group along the way. Their material can be changed to be bouncy or sticky, allowing them to fall great distances and bounce or to stick to surfaces to move up and around obstacles. Mar can also be changed into other shapes in addition to the material changes, moving between balls, blocks, and triangular ramps. Each level is structured with an entrance and exit, with platforms and pits that the player must help the mar navigate.
Implementation was carried out by a team of 2 over a period of 7 weeks, with 1 week for initial concepts and design though the design process continued throughout the development of the game.
Space to Fly
Instructions:
Press space to fly upwards
If you aren't flying upwards, you fall
Avoid hitting the green and blue
Pressing P will pause the game
This was an experiment to learn how to use html canvas elements. I'm still using this game and a few other simple tests to experiment with the best way to construct a JavaScript game engine (or at least the best design for me to work with). For working code, I'm moving towards a WebGL drawing component so that I can more easily test shaders and effects, but I've preserved this iteration as an example of one of my quick web prototypes.
ALT
ALT
ALT
