Hello, I'm a student at the University of Washington majoring in Computer Science and minoring in Physics. I have a strong interest in speech modeling with deep learning, web applications, and generally anything to do with computing.
I built the database commenting feature. Previously, you could only comment on an entire Notion page, specific text selections, or entire Notion blocks so this was a widely requested improvement. Here's a Demo.
Database docs | LinkedIn postI helped build parts of an inference system that ran DINOV2 on an IOS device in order to give a wearable ring the ability to control smart home devices. Learn more by reading the paper.
PaperI performed accuracy and performance tests for ONNX Runtime on next generation CPUs in order to identify, record, and analyze flaws and regressions. This involved having to make test harnesses and low-level automation scripts that ran batches of the latest vision, text, audio, etc. models across many different systems. The tools and automations I created greatly sped up the testing processes for my team and were distributed to partner teams.
I fixed performance issues with a layout/rule engine used by the Amazon.com product detail page that existed for 3+ years. Specifically, an internal tool used to make changes to product detail page layouts experienced loading challenges which I fixed by moving heavy front-end procedures to backend cloud services while also restructuring processes to be asynchronous. This involved writing multithreaded JSON processing algorithms in Scala that handled large (over 60 MB of data) query responses to a tree-like database.
I designed and implemented a refund feature for rentable phone lines. WestBold provides anonymous phone lines for online verification purposes but was in need of refund functionality. I developed the backend service in C#, involving numerous security features that avoided fraudulent use, utilized Entity Framework Core in order to build on an existing PostgresSQL database application, and created public API endpoints along with UI additions that allowed the service to be accessed multiple ways. UI features were created using the Blazor framework.
API docsTrained a Dense neural network to identify whether a movie review is positive or negative based on its text. Utilized the Keras deep learning API along with TensorFlow to train the model on a GPU using google Collaboratory. The model was able to achieve ~87% accuracy on test data and was trained on data from the Keras IMDB data set.
Demo | Github LinkDeveloped an android game using the unity game engine and C# scripting. The game included in-App purchases and advertisement using the Unity IAP package and Unity ads SDK respectively. The game was released to the google play store.
Google Play StoreCompeted in a two-day hackathon with a partner to create a website that served covid data. Worked primarily on the back end where I built a Node.js server which constantly refreshed the data by collecting information from a CDC website using a python web scraper.
Hackathon listing