Selected Data Science Projects for Harrison Yu

---

Resume Matcher Pipeline Using Sentence Embedding

Keywords: Big Data, Pipeline, Spark, no-SQL, Airflow, Cloud Computing(GCP), Sentence Embedding

• Designed a data jobs recommendation platform for given resumes using Airflow, MangoDB and Spark ML
• Experimented with different embedding models using Sentence Transformers to optimize accuracy of the system
• Built a pipeline to collect job data automatically and store over 3500 job descriptions in NoSQL database

Browsing Time Minimization for Streaming Services with Experiments

Keywords: A/B Testing, Experiments, Hypothesis Testing, Response Surface Methology

• Minimize browsing time by ~40% by conducting multivariate experiments to find optimal design factors
• Run partial F-tests to determine interaction effects between design factors and fitted a response surface model to further locate the optimum

NYC Household Broadband Adoption Rate Analysis

Keywords: Geospatial Analytics, Regression Analysis, Correlation Analysis, GIS, Open Data

• Conducting a regression analysis to explore the digital divide in NYC neighborhoods on internet adoption rates and socioeconomic factors in ArcGIS
• Improved R-2 score by ~30% by implementing a geographically weighted regression (GWR) model
• Visualized clusters and outliers of internet adoption using spatial weight matrix and Local Moran’s I in Python