Hi! I'm Zeyi Li
Current Student at
UC Berkeley
I'm currently pursuing Data Science with a minor in Computer Science.
Relevant courses I have taken so far: CS61A (Structure and Interpretation of Computer Programs), CS61B (Data Structures and Algorithms), CS70 (Discrete Math and Probability), Data C8 (Introduction to Data Science), Math 54 (Linear Algebra and Differential Equations).
Courses I am taking now: CS188 (Introduction to Artificial Intelligence), Data C100 (Principles & Techniques of Data Science), Data C140 (Probability for Data Science)
P.S. refresh the page for a cool surprise
Projects
Movie Rental Application with Recommendation System
Designed database for users, movies, rentals, and ratings using MySQL Developed frontend with React (Node.js) to browse movies, viewing details, and renting/returning Used Python and Flask to create REST API for user registration and rental system for backend Created recommendation models using popularity-based, content-based, and collaborative filtering techniques through Pandas and Scikit-learn
Redthread
Developed hybrid e-commerce and social media mobile application using Flutter and Stripe that allows users to buy clothing items tailored to fashion style preferences and post outfit boards, designed with Figma Used Firebase Authentication for managing user accounts, Firestore to store and sync user data, posts, and product information, and Cloud Messaging for notifications Implemented content-based and collaborative filtering recommendation system based on cosine similarity, using Python and Scikit-learn to vectorize fashion styles and user interactions with post attributes Created Python scripts with Shopify API to automatically store product data from different stores into database, and smtplib to send order confirmation emails to sellers
Stock Price Predictor/Advisor
Used Python libraries NumPy and Pandas for processing and formatting stock data from yfinance API Developed Linear Regression and Support Vector Regression models with Scikit-learn, used grid search technique to optimize SVR model parameters Created Long Short-Term Memory model using TensorFlow, used sliding window technique to split test and train sets Visualized past 30 days' price data and predicted values with Matplotlib Developed frontend using React (Node.js) and Axios for HTTP requests, used Flask for backend services
Experiences
01
JP BioInfo (June - Sept 2022)
JP BioInfo is a bioinformatics company that uses machine-learning methods to predict molecular reactions, whether it be in drug development or disease predictions. Using openCV and Pytesseract, an OCR (Optical Character Reader), our team developed software to extract data from pdf, jpeg, jpg, and png file types and transfer them to CSV files. However, many characters were misread, such as 0 mistaken for O, or 5 for S. I then used pandas to iterate through the data and clean what could be corrected by analyzing trends, reducing overall misreads by 70%. Finally, all this data was moved to Excel sheets. I used Scikit-learn to build a KNN model that used 10+ molecular attributes to predict brain and plasma concentration in mice.
Scroll to see more →