Projects and Experience
Some of the things I've been working on lately
-
Chefin
Cooking app that generates recipes based on a user’s available ingredients, optimizing to use as many inputs as possible while maintaining realistic cooking constraints. Built with Next.js and Express.js, with ingredient and recipe persistence in MongoDB. Designed with RESTful APIs in mind for ingredient management, recipe generation, and saved recipes. Currently experimenting with and evaluating multiple different recipe generation APIs, comparing output quality, ingredient coverage, and latency. Styled using shadcn/ui.
github try it out! (coming soon) -
Semiconductor Market Prediction
Collaborated in a team of five to analyze historical semiconductor industry and financial market data to predict short-term and medium-term market trends. Designed, trained, and evaluated three different machine learning models, comparing performance using metrics such as RMSE. Focused on feature engineering from time-series financial data and assessing model robustness under volatile market conditions.
market prediction writeup -
Computer Vision
Collection of projects exploring computer vision from first principles, spanning image formation, image processing, multi-view geometry, and modern generative models for UC Berkeley course CS180. Implemented core vision techniques including image filtering and convolution, Fourier analysis, image pyramids and blending, homographies and image mosaics, and automatic image alignment.
cs180 projects -
Perceptrix
Augmented-reality glasses try-on app that detects a user’s face shape and recommends compatible eyewear styles. Built with React and Expo, using TensorFlow.js for face detection and face-shape classification. Implemented real-time virtual try-on using AR overlays. Worked in a team of five using Agile methodology over eight weeks, contributing to both frontend development and ML integration. Presented the final product to ~200 attendees at UC Berkeley’s Jacobs Hall of Design.
github -
Football Analysis System
Real-time football analytics system leveraging computer vision and machine learning to track players, referees, and the ball during live gameplay. Trained YOLOv8 on Bundesliga match footage for object detection, applied K-means clustering to group players by team, and used optical flow to track player movement and speed across frames. Planning to revisit with personal game footage, hoping to bring professional-like statistics to my own game.
-
Artificial Intelligence
Collection of hands-on projects for UC Berkeley CS188 exploring intelligent decision-making, planning, learning, and reasoning under uncertainty. Projects ranged from search and game-playing agents to probabilistic inference systems and classical machine learning models implemented from scratch.
github -
Semester Abroad @ Hong Kong University of Science and Technology
Studied database systems and completed a capstone project optimizing an internal student-faculty project matching platform. Improved query performance by 3× through indexing optimization, query restructuring, and eliminating inefficient joins. Gained hands-on experience with performance profiling and real-world SQL optimization.
-
Gitlet
A simplified version control system inspired by Git, supporting core version control features such as repository initialization, staging, committing, branching, merging, and checkout. Built in Java, using underlying data structures to efficiently store commit graphs, file snapshots, and branch pointers, using multiple different algorithms to traverse commit graph and support git functionality.
github