Duke University

I am currently a research assistant working on an AutoML project for assisted learning. The objective of the project is to automate the entire machine learning process-- from data preprocessing to hyperparameter optimization -- for gradient assisted learning, which is a novel method of using a decentralized system of models that can improve each other without leaking data to each other. My specific role relates to exploring different options for hyperparameter search, particularly grid search, gradient-based optimization, or automated bayesian optimization.

I've also served as an undergraduate teaching assistant for a year. I helped teach the classes CS330: Design and Analysis of Algorithms and CS201: Data Structures and Algorithms in the Spring 2023 and Fall 2022 terms respectively. Some of my duties included grading the works of students, which consisted of projects, proofs, and algorithms, holding weekly 2-hour office hours to answer questions regarding concepts, and leading weekly discussion sections, in which I and a partner would help our class of 25+ students go over the material from the past week.

Capital One

The overall project I worked on was a feedback form component for an internal call quality platform called SAGE serving around 16,000 monthly users. Although it seemed like a small task at first, considering it just consisted of two sparse pages with very few inputs -- one for the actual form itself and one for certain people to query past forms -- I quickly learned it was not. The project took all 10 weeks, during which we:
• designed the components using Vue.js
✧ built a validator for the inputs and passed in the data as a JSON to a backend-for-frontend layer that we also made
• registered entitlements that would only make certain users see the search option
✧ created search and submit endpoints on an AWS lambda that received the Information from the BFF layer once passing through a security layer
• built unit and functional testing with almost 100% code coverage
✧ submitted the HTTP requests to AWS OpenSearch, which contained indices that we also configured
✧ created a pipeline and deployed the application to qa then prod

Whew! It was a lot of fun, and I learned so much from everyone around me, especially my fellow interns. My specific responsibilities were the star bullets, but we all helped each other often.

Cigna

In the beginning of my internship, I worked on creating dashboards for a web-app called the Cigna Global Fitness Challenge, which promotes fitness campaigns that employers could subscribe to. I utilized Google Analytics to log user interactions and a Google Firebase Dashboard to display these stats in realtime. Then, I created an AWS Cloudwatch Dashboard to monitor the health of the MySQL database in AWS RDS and the AWS EC2 instances used. Together, these dashboards helped prevent annually-occuring scalability issues by a significant margin because developers were now able to predict issues better and react quicker.

Then, I briefly worked on creating a web scraper in BeautifulSoup for visualizing Cigna's employee hierarchy as a side project for my manager. Although the task wasn't too hard, scraping all the data and storing it in a SQLite DB took extremely long. Thus, with another intern, I multithreaded the process using Python's threading library (the Global Interpreter Lock did not matter much here because the program was bottlenecked by I/O-bound tasks), reducing the time from ~6 hours to ~30 minutes.

I spent the last half of the internship working on revamping a virtual reality meditation app. I and another intern worked in Blender, Unity, and C#, utilizing the Meta Avatars SDK and the Oculus Integration SDK, to change the existing app, in which a user could only look around at a calming scene, to include a social mode. In this interactive multiplayer mode, users could move around, talk to others, choose avatars, and grab and throw objects. I primarily worked on the latter two capabilities, and I was delighted to hear our work was met with so much enthusiasm that my team presented it at Cigna's booth at the Fall '22 Grace Hopper Conference!

BNSoft

I was located in Seoul, South Korea during this internship, and I really got to experience how widely work cultures can range. Most of my time was spent creating proof of concepts like leader-lines and minimaps (some of which actually ended up being used!) in Vue.js to be used in the product, which was an application for building AUTOSAR-based operating systems. Hyundai took a strong interest in the overall product and decided to use it in their development process, which was really cool to see.