I am a determined student studying NLP at University of Waterloo advised by Yuntian Deng. Simultaneously, I am also nearly a professionally credentialed P&C Actuary (insurance engineer) with 6/7 exams fullfilled for associateship with the CAS. My research interests are in AI and ML. Right now is a very exciting time for AI with the advent of AI for everyone via ChatGPT. In the midst of the excitement I have identified one important area I wish to research, which is communication between LLM's. I believe communication frameworks are critical to enabling meta-learning capabilites.

I also enjoy bbq/smoking, basketball, pool, finance, golf and am a huge AMD fan in their pursuit to make AI GPU compute open and available to anyone.

Selected Projects

	Morse Code Transformer I implemented the transformer architecture from scratch. I trained two LLMs, one on english characters and another on morse encoded english characters. I found that increasing ONLY the context length proportionally yielded similar performance between the morse code and english LLM.
	Multi-Implicit Chain of Thought Reasoning via Knowledge Distillation I expanded upon a paper by training the LLM to perform simultaneous 2x2 inference using the same knowledge distillation process all packaged up in a nice and easy to use notebook.
	ChatGPT vs Keyword ML I trained a NN, RNN and GBM to predict amazon review scores from keywords and compared to ChatGPT using the openAI API. The results demonstrated that ChatGPT understands context by significantly outperforming traditional keyword models in binary score prediction.
	Zoomerified Java Code Compiler This is just a silly project I did for a course. I created a Java program that emulates a compiler. The emulated compiler takes in "Zoomerfied" Java structured code and parses, compiles and runs its translated java program.