S
Szu-En Tu
About
Detail
California, United States
Résumé
Jobs
verified_user
0% verified
Data ScientistQventusMar 2025 - Jun 2025 (4 months)• Developed and deployed a production-grade XGBoost model on AWS SageMaker using 1.2M+ surgery records to predict last-minute cancellations, reducing delays by 20% through real-time scheduling alerts. • Engineered 20+ time-sensitive features from EMR and operational data in Snowflake, carefully controlling for data leakage and boosting model AUC by 8%. • Built a modular ETL pipeline using PySpark SQL to process 580K+ records weekly and applied SMOTE to handle severe class imbalance, improving minority-class precision by 12%. • Deployed real-time inference endpoints with integrated monitoring and logging, enabling clinicians to make scheduling decisions with live model output embedded into hospital workflows.- KData ScientistKlarity HealthOct 2024 - Mar 2025 (6 months)• Created a logistic regression model with a 96% capture rate to predict provider booking likelihood, streamlining targeted outreach and improving sales conversion efficiency. • Performed geospatial market segmentation using K-Means clustering and regional socioeconomic datasets, identifying the top 20% growth regions for expansion and targeted ad campaigns. • Led iterative A/B testing cycles and KPI tracking in Looker Studio, optimizing the provider sign-up funnel and raising booking conversion rates from 0.4% to 1.25%. • Conducted root-cause analysis of booking drop-offs via time-series modeling and visibility scoring, reducing spam-related losses by 77% through ranking algorithm tuning.
Education
verified_user
0% verified
- M.S. in Data ScienceUniversity of San FranciscoJul 2024 - Jun 2025 (1 year)
- B.S. in Data ScienceUniversity of San FranciscoAug 2020 - May 2024 (3 years 10 months)
Projects (professional or personal)
verified_user
0% verified
- EEduGgenieJan 2025 - May 2025 (5 months)• Built a conversational RAG (Retrieval-Augmented Generation) assistant using Gemini API and Streamlit to answer study-related questions based on users' uploaded class materials and personal calendars. • Deployed on Streamlit Cloud with automated CI/CD pipelines via GitHub Actions, enabling seamless updates and zero-downtime maintenance without the need for custom cloud orchestration.
- SScalable Crime Data ProcessingJan 2024 - May 2025 (1 year 5 months)• Processed over 8.9 million NYC crime records using PySpark on AWS EMR Serverless, demonstrating scalable, distributed data handling for real-world datasets. • Built an end-to-end distributed ETL pipeline that ingested, cleaned, and transformed raw incident-level data stored in S3, delivering high-performance insights for downstream visualizations and analysis.
- VVisualizing Homelessness TrendsJan 2005 - Jan 2013 (8 years 1 month)• Analyzed nine years of national homelessness data revealing a 15% increase in overall homelessness, with key disparities by region and policy environment. • Designed a D3.js interactive tool to present data on homelessness trends, increasing engagement by 25% and guiding equitable housing discussions in top-growth regions.