My Projects

---

• TrashGPT:
  ^{I collaborated with three peers on a semester-long project to develop a model that could generate podcast transcripts mimicking the hosts’ style and voices. By fine-tuning language models and training a vocoder, we successfully created a system that could generate new podcast episodes, showcasing the potential to revolutionize podcast production through AI-generated audio content. In the future, we aim is to refine and advance this pipeline by automating data labelling.}

  ^{Skills: Python, Bash, NLP, Pytorch, Data Preprocessing, Model Fine-tuning, Hyperparameter tuning, Audio Synthesis}

• DLSS For All:
  ^{As someone who spends a lot of his time gaming, and having recently gotten an RTX GPU, I was very impressed by Nvidia’s DLSS and wanted to implement something similar. This project is exactly that, after having gained some knowledge from a few graduate Machine Learning courses, I wanted to train and implement an upscaling algorithm from scratch. I was very much satisfied with the results, however when I tried to use this model to upscale videos, it was too slow. Some future work with this project can involve making a more efficent model that is able to upscale videos in a more reasonable amount of time.}

  ^{Skills: Python, Tensorflow, Image Upscaling, Data Preprocessing, Model Optimization, Hyperparameter tuning}

• Worth It or Not:
  ^{This project was a way for me to try and estimate my enjoyment of a game that I’m considering to potentially buy. By scraping data about the game’s genre and estimated playtime, I developed an automated ensemble machine learning model that can predict user’s satisfaction in a game they may be intersted in. From there I let a few of my friends (and myself) test the software and worked on making improvements such as remembering user queries, improving the search feature to be an actual search engine, UI, etc.}

  ^{Skills: Python, Tensorflow, SkLearn, Automated Machine Learning, Web Scraping, CICD, Search Engine}

• Mastermind AI:
  ^{As a kid, one of my favorite board games to play was Mastermind. If you’re not familiar with it, it’s a 2+ player game where one player puts together a code (4 different colors from a pool of 8) and the other player(s) has to guess the code within a certain number of steps. With each guess, the person who put together the code tells the code breaker how many colors are in the right place, and how many colors are correct but in the correct slot. This project was inspired by a sudden spark into the game of mastermind after having found the game in my basement after moving, which inspired me to make an AI for the game.}

  ^{Skills: Python, AI, Command Line Interface, Algorithm Design, Game Theory}

• Cifar 10 CNN:
  ^{Before taking formal courses, I delved into machine learning as an enthusiast, challenging myself by exploring the Cifar 10 dataset with a convolutional neural network. Achieving a 90% accuracy on the test set, this endeavor taught me the importance of data preprocessing for optimal results. I now look forward to tackling more demanding datasets like Cifar 100 and ImageNet to further enhance my skills and understanding of machine learning.}

  ^{Skills: Python, Tensorflow, Deep learning, Image Classification, Data Preprocessing, Model Optimization, Hyperparameter tuning}