Shiyu Sun

• DSCI 551 Foundations of Data Management • Led and managed a team of TAs, delegating grading tasks, coordinating logistics with the professor, and proactively resolving challenges to optimize course operations. • Developed and implemented PyTest auto-grading scripts on AWS EC2 to streamline and enhance the efficiency of SQL and Spark assignment evaluations, ensuring consistent and accurate grading for ~250 students. • Supported students with debugging and understanding course concepts, including SQL query optimization, big data analytics (Spark, Hadoop), cloud computing (AWS, BigQuery), and databases (MySQL, MongoDB, DynamoDB).

• Research Topic: NLP-Driven Data Extraction for Analyzing Hospital Medical Debt Collection Practices in Washington • Utilized web scraping with Selenium to download 30,000+ civil case files, performing data preprocessing to validate and prepare the dataset for for subsequent NLP tasks while ensuring data integrity. • Applied Active Learning and SpaCy for Named Entity Recognition (NER) to extract medical debt-related entities from mixed-layout court documents with handwritten text, using iterative training for improved extraction accuracy. • Enhanced extraction accuracy by 18% using LayoutLM for improved document parsing and information retrieval, empowering public health researchers with deeper insights into the topic.

• Hold weekly workshops on film topics providing a community for student filmmakers • Manage student-led film productions allowing students to work on short films gaining hands-on experience with other student creatives

• Engineered and fine-tuned regression and tree-based Machine Learning models in Python, using over five years of sales data to evaluate ad spend impact on sales, enabling better budget allocation and a 10% increase in campaign ROI. • Combined insights from both SARIMA and Prophet time-series models to forecast sales from trends, seasonality, and cyclical patterns guiding inventory management and targeted sales strategies like promotions or discount. • Automated digital marketing (web traffic) reports with SQL and Tableau, focusing on SERP positioning, click-through rates, and bounce rates, informing SEO and SEM strategy adjustments, boosting conversion rates by 12%. • Delivered bi-weekly data-driven reports to stakeholders, conducting competitor analysis on online sales, pricing and promotional strategies, and market share trends across e-commerce platforms, providing insights that optimized e-commerce strategies and drove ROI for key clients, including L'Oréal and LEGO.

• Analyzed influencer survey data weekly using Pandas and Excel, targeting engagement rates, follower demographics, and cost-per-collaboration to support the operations team in forming partnerships and optimizing campaigns. • Designed and implemented ETL pipelines with Hadoop to regularly process and organize real-time raw user and sales data into a clean, structured database, improving data accessibility and enabling more efficient analysis. • Utilized SQL and PySpark to analyze large-scale user and sales data, focusing on metrics like customer acquisition cost, churn rate, and revenue per user, providing insights to enhance customer lifetime value and optimize conversion rates. • Developed interactive dashboards to monitor product and marketing KPIs, presenting insights to stakeholder weekly, including product managers and sales teams, to drive data-driven decision-making and strategy refinement. • Conducted A/B testing on the platform's homepage to optimize new users’ first-order conversion, evaluating category recommendation algo with segmentation and statistical analysis, resulting in a ~10% increase in conversion rates.

• Dean’s List: 2020-2021, Summer 2021, Winter 2022, Spring 2022