• Analyzed entertainment ticket sales data from Snowflake, performed customer segmentation, and created Python visualizations to identify high-value customer groups for strategic marketing decisions• Automated ETL pipeline for Looker dashboards using MySQL and Python, monitoring over 15 models' performance metrics, such as mean absolute error and custom-defined metrics, to detect model drifts• Developed a machine learning model using Random Forest to predict ticket sell-through; recommended pricing based on predicted sell-through and business rules, resulting in an approximate $100k increase in fall profits• Conducted experiments using MySQL to query six feature sets from multiple databases and Python for feature engineering presale pricing models, resulting in a 2-5% increase in ticket sales• Documented projects process; shared model performance metrics and business insights with cross-functional teams

Established classification models to detect, measure, and classify influential media posts containing misinformationPerformed KMeans and CNN to identify text screenshots from 1million+ images on a serverPresented weekly results and collaborate with a team of 8 people to generalize models to multicultural settings to identify bad actors’ traits and misinformation diffusion through text recognition and analysis

Analyzed and screened companies' data to decide the investment amount. Gave presentations and conducted interviews with TSVC leaders who work in the machine learning industry. Wrote articles of companies that focus on data solutions and publicized their innovative strategies and company design.

Title: Faster Solver For Multiple Linear Systems Via Block Conjugate Gradient. Improved a fast iterative solver for multiple linear systems. Updated conjugate gradients method to preconditioned block conjugate gradients (PBCG) method. Implemented four algorithms in Python and tested the speed of PBCG under fluid flow contexts. Increased the speed to find the solution of multiple linear systems 3X faster. Presented the final results with visualization to the faculty of Courant Institute of Math at New York University.

Worked as a tutor for Discrete Mathematics and Linear Algebra classes. Communicated withProfessors on following math standards. Answered math problems from students. Provided helpful feedback to students.

Title: Trace Reconstruction: Recovering the Original Message with High Probability.Consider that there is a random binary sequence. Every time there will be a new trace when the original random binary string is transmitted through a “deleting channel,” which extirpates bits arbitrarily with a random probability. Hence, we want to use distinct subsequences to recover the original binary string/sequence with high probability.Applied binomial probability, Hoeffding’s inequality, and Sterling’s formula to reconstruct the broken sequence. Implemented two algorithms using Python and tested accuracy with data larger than 100 bits.

Math tutor for University Neighborhood Middle School, New York. Taught middle school students algebra, trigonometry, and other math concepts. Communicated with teachers and parents to improve students’ academic performance.