Loading
Welcome To My Portfolio

BOSUNG.H

PORTFOLIO

About Me

01

Hi!
My name is Bosung Hwang.
I love playing sports and working out.

Education

University of California, Berkeley
Bachelor of Arts in Data Science, Senior
Cumulative GPA: 3.71/4.00

Relavant Coursework

Computer Science & Engineering
CS61A: Program structures
CS61B: Data structures
CS61C: Computer Architecture
CS170: Efficient Algorithms and Intractable Problems
EECS127: Optimization Models in Engineering
Math110: Linear Algebra

Machine Learning & AI
CS189: Introduction to Machine Learning
Info159: Natural Language Processing

Data Science & Engineering
Data100: Principles & Techniques of Data Science
Data101: Data Engineering
Data140: Probability of Data Science

My Works

02

Campass

Campass is a cross-platform campus discovery app that helps students find and explore places in real time. I launched the MVP with 25+ UC Berkeley students, building an API-driven ingestion pipeline and a Firestore-backed REST layer aggregating data on 500+ campus locations. I integrated an AI-powered campus agent with schema validation, OpenAI function-calling, and robust error handling to support real-time Q&A. In preparation for expansion, I positioned Campass for a multi-campus rollout through the UC Berkeley SkyDeck 2025 cohort.

ReactNative
Node.js
Express.js
FireBase
External APIs

SnackHub

SnackHub is a gamified food ordering app designed to optimize cooking operations for small businesses while making the customer ordering experience more engaging through gamified and social features. Qualified for the Naver D2 Startup Factory First Round, I led a 3-person team to build core functionalities including a reward-based ordering system, interactive UI, and personalized restaurant recommendations. I developed REST APIs on Google App Engine to manage user data and order flows, and piloted the app with a local restaurant, gathering feedback that led to a 60% improvement in user experience.

ReactNative
Node.js
Express.js
Google App Engine

Wikipedia Talk-Page Comment Classification

Built an end-to-end NLP pipeline to categorize editorial feedback on Wikipedia talk-page comments into three tiers: minor edits, moderate suggestions, and major disputes. Fine-tuned a BERT model in PyTorch, achieving 87% test accuracy and outperforming baseline methods by 22%

PyTorch
NLTK
scikit-learn
NumPy

Yelp Data Analysis

Analyzed real-world Yelp data to uncover insights into business attributes, user behavior, and review patterns. Developed queries in PyMongo to filter businesses, aggregate reviews, and extract nested fields from JSON-like documents for efficient analysis

MongoDB
PyMongo

Boba Time

image of BobaTime homescreen

Boba Time is an iOS application developed using Swift and SwiftUI/UX. The app features an engaging series of quizzes that utilize a custom algorithm to recommend a personalized boba drink. Users can save their favorite boba selections to a remote database powered by Firebase. The application was recognized as the winner of the UC Berkeley App Cubstart Competition.

Swift
SwiftUI/UX
FireBase

Ehyun Elementary School English Musical Website

image of ehun website

Ehyun Elementary School's English Musical Club is a state funded club that has 50+ members every semester. I created their official website using various tools and libaries.

Website
HTML/CSS
Javascript
React

Experience

03

Research Assistant - Full Stack Development

UC Berkeley Sky Computing, CA
- Contributed to the development of Chatbot Arena through backend optimizations and improvements of UI/UX, enhancing user engagement by 20%
- Supported the Arena-Hard benchmarking pipeline by connecting backend LLM evaluations with frontend components for real-time comparisons
- Developed data collection APIs and integrated user interaction metrics into a live benchmarking dashboard using Next.js
- Conducted design and code reviews, and debugged performance issues across API endpoints

Next.js
Postgres
Tailwind
ShadCN

Software Engineering Intern

image of ehun website

Karrot - Search & Discovery Team, South Korea
- Extended SQL-backed API endpoints to retrieve user interaction data for search and filter features, improving data delivery efficiency on internal tools by over 30%
- Analyzed 1M+ activity logs to identify user behavior patterns for personalization models
- Contributed to frontend layout fixes and styling adjustments for the new Karrot homepage prototype
- Collaborated with engineers on full-stack feature rollouts, debugging issues across the data and UI layers

Python
SQL
React
Node.js

Data Science Internship

San Diego County Taxpayers Association
- Automated extraction of public expenditure data from government websites, reducing effort
- Created reports for data-driven policy recommendations on housing, taxation, and budget allocation

Requests
BeautifulSoup
Pandas
Seaborn

Undergraduate Course Staff/Tutor

Principles & Techniques of Data Science (Data 100), UC Berkeley,CA
- Led weekly discussion sections with 30+ students and held office hours and group tutoring sessions

Contact Me

04

Thanks for stopping by!

Have any questions, comments, or want to collaborate on a project? Let's have a talk!

BOSUNG.H

PORTFOLIO