hello! welcome to this projects page! there's not an insane
amount of stuff in here, but i have done a couple of
pretty cool projects that i wanted to put on this page.
if you're interested in learning more about any of these projects, feel free to reach out to me. alternatively, there may be more info on other pages on this site, or you can refer to my github page! while not all the resources may be in there, you might find it's worth checking out. sometimes, not all the code or data for a project is immediately available. i'll try to update this page and my other pages accordingly.
thanks for checking this page out! i hope you enjoy and find these projects cool!
president botrick is a project i started in my senior year of high school in 2021 in a class called asi, or advanced scientific investigations. the aim of the class was to complete a year long project on anything scientific, although most projects were related to biology, chemistry, physics, or computer science.
i decided to start the project because at the time, i was curious about how natural language processing could play a role in politics. the name of the project actually comes from rodrick from diary of a wimpy kid, i thought it would be funny to name it after him, and so it's a mashup of that and 'robot'.
the project itself involved two parts: analyzing political speech based on the reaction it receives from an audience as well as generating new political text with the understanding of "influential" speech already developed. this project used the corps dataset, a corpus of thousands of political speeches from famous figures in history. the beef of the project is contained in the usage of gpt-2 and bert, both fine-tuned for the data and task accordingly. bert takes control of the first portion, assisting a binary-classifying neural net in deciding if a line of a speech is "impactful" or not. gpt-2 acts as the generative portion, training on those "impactful" lines that the first part spits out, and writing something it thinks may warrant a similar reaction or decision from a model or audience.
this project ended up being written to a paper and published at aaai 2023 workshop for ai and diplomacy alongside professor julia hirschberg from columbia university!! very cool!!
| code | paper |
music and ai
this project was one of my first ai projects back in 2020. the pandemic locked us all indoors and i decided to start playing around with some models and seeing what i could do with them. i ended up deciding to use it for some music project, although more related to symbolic representations of music and lyrics, rather than working with actual audio.
this project is based around using ai to classify and generate music by its genre, lyrics, and midi files. it's a very simple project that relied on some transformer-based models to do most of the work. it's not incredibly technologically complex, but it was an exciting thing to learn about.
there is a write-up available for this project, at the same link as the code, that i would attempt to legitimately publish if this work didn't violate many, many, many...., many copyright laws. but alas, it's nice to have all of the information on the project in one condensed place.
| code |
updating soon...keep an eye out