about me
My first encounter with code was in third grade playing around with BonsaiJS, a graphics library that allowed me to animate shapes. I had a lot of fun making bubbles follow my cursor, and I've enjoyed coding ever since!
Now, I'm a third year at UC Davis, double majoring in Computer Science and Statistics on the Machine Learning track, and minoring in Technology Management.
I've worked as a software engineering intern at Apple on the Shazam team, where I had the privilege of delivering production code for a client-facing feature and creating a functional prototype for a research project.
I've also had the opportunity to work on several engaging and fulfilling projects while volunteering as a Technical Director at HackDavis and a Lead Software Engineer at #include.
Technologies I use
General
- PostgreSQL
- MongoDB
- Unit Testing
- Authentication
- Multithreading
- Docker
- Git
- Linux/Unix
- RAG
- CUDA
Languages
- C++
- Swift
- Python
- Golang
- Javascript
- Typescript
- Java
- HTML & CSS
- SQL
- GraphQL
- R
Libraries/Frameworks
- SwiftUI
- Next.js
- PyTorch
- React.js
- React Native
- Apollo GraphQL
- Node.js
- Playwright
- Auth.js
- LangChain
- Unity
In my free time, I love weightlifting, cooking/baking, crocheting, and exploring the great outdoors!
experience
Software Engineering Intern @ AppleJune 2023 - Sept 2023Engineered the Music Recognition module in macOS Control Center for the Shazam team using SwiftUIIndependently started a digital signal processing research project and presented a product prototype to Shazam's directorAddressed issues and built features in the Events app and macOS Control Center, benefiting over 100 million monthly usersDemonstrated initiative by pitching a sleep apnea detection prototype to 5 members of the Health team leadership
projects
An iOS social media app in which users can use ShazamKit's music recognition to identify songs and make posts. Users can add a caption and location to their posts, as well as see each other's posts in either a list view or a map view. From the posts, users can also play the songs that others have Shazamed through Apple Music. The posts are stored in a MongoDB Atlas instance.
SwiftUI, Realm, and MongoDBTrained a deep convolutional U-Net to isolate vocals from songs. We converted raw audio to spectrograms and trained a mask spectrogram that is applied to the input song to get the vocals. I'm currently implementing a transformer variant of the U-Net and deploying it using Flask.
PyTorch, Scikit-Learn, and LibrosaThese projects were part of an operating systems course at UC Davis. I implemented a distributed file system similar to AWS S3, a multithreaded web server capable of handling multiple clients simultaneously, a user mode Unix shell with file redirection and parallel execution, and several Unix utilities.
C++ and Shell ScriptThese projects were part of a graduate course on advanced deep learning at UC Davis. In one of the projects, I made several modifications to the nanoGPT code created by Andrej Karpathy in order to improve performance and runtime. The modifications include reducing dimensionality of key, query, and value vectors, implementing sliding window attention, and changing the MLP architecture. The other project involved performing knowledge distillation in order to improve a small pretrained vision transformer model with the help of a larger model.
PyTorchAn iOS shopping list app in which users can scan price tags with text recognition and track their purchases. Item prices are stored in a MongoDB Atlas instance and made available through a REST API created with Node. Stores can call this API see which items users are buying at which stores and make business decisions with this information. Users can decide where to shop based on logged prices.
SwiftUI, Node.js, Express.js, Realm, and MongoDBA website leveraging the Spotify API allowing users to find people in the same area with similar music interests. Currently, users can pick an area and see what songs people are listening to live. In the future, we plan to display a world map where users can drop a marker anywhere and see what people are listening to there.
React.js, Express.js, Apollo GraphQL, PostgreSQL, and MaterialUIAn iOS app in which users can find others traveling in a similar direction and carpool together. Matching users is done on a backend server made accessible through a REST API created using Node and Express. Users are alerted through Pusher's publish and subscribe framework, and user data is stored on a MongoDB NoSQL database.
SwiftUI, Node.js, Express.js, Pusher, Realm, and MongoDB