Leo Li

• Developed and optimized data pipelines for model training, translating complex business insights into actionable features. • Scraped and cleaned data from diverse sources, including crypto and sentiment data, enhancing model efficiency. • Collaborated with cross-functional teams to integrate data workflows into the algorithmic trading engine, improving performance.

• Built an automated ETL-based address matching pipeline to monthly ingest 150,000+ transaction records from PostgreSQL to AWS OpenSearch • Parsed and normalized addresses using usaddress, and computed confidence scores based on component-level differences and rapidfuzz similarity. Enabled scalable exact and fuzzy search via OpenSearch • Scheduled batch ingestion with cron and logging, improving monthly matching coverage from 49% to 90%, with zero mismatches on confidence scores ≥95

• Developed an end-to-end NLP analytics pipeline to analyze over 10,000 Reddit comments from NBA team subreddits, measuring fan aggressiveness and neutrality toward others based on game outcomes (win/loss/neutral) • Boosted processing efficiency by 30% using AWS S3 for storage, AWS Glue for ETL orchestration, and Redshift for querying team performance data alongside fan sentiment • Designed modular preprocessing features for future scaling to Spark, Hive, and Flink environments, supporting broader data warehousing and real-time analytics use cases

• Built a full-stack web-based clustering and scoring platform using React.js (Next.js), Node.js, MongoDB Atlas, RESTful API • Support performance evaluation of 100+ teachers and 40 regional managers at a national charity foundation • Reduced k-means within-cluster variance from 60% to 40% by implementing automated data preprocessing (boolean conversion, Z-normalization) and integrating SQL-based data manipulation • Enhanced clustering accuracy and score precision by expanding to 20+ domain-informed subgroups, applying business rules and data segmentation strategies to refine volunteer performance metrics

• Assisted in designing over 4 marketing plans and designed the marketing campaign for the new financial app • Increased download rates by over 80%, raising the number of downloaded records from 16 to more than 30 • Assisted in developing marketing strategies and gaining insights into departmental operations, which significantly improved the department’s efficiency, reducing the average development time from 10 workdays to 3

•Developed R Shiny web software to automate statistical workflows and data visualization, enabling 200+ faculty and students to conduct tests and interpret results without coding •Improved user experience by over 20%, raising satisfaction scores from 75 to 90 and increasing usability by 10% through optimized app design and functionality •Boosted development efficiency by 40% by standardizing R Shiny code with Tidyverse, replacing outdated functions and cutting delivery time by 7+ days. •Built an all-in-one variance testing software (Bartlett, Levene, Fligner) with integrated summarization, visualization, and hypothesis testing for non-technical users

•Organized and coordinated volunteers to teach Chinese language and culture to elementary school students •Designed Chinese teaching courses in collaboration with another coordinator