Machine Learning Data Analyst

MOSHU HUANG

*****.*****@****.*** • 919-***-**** • Raleigh, NC • Linkedin.com/in/moshu-huang

EDUCATION

DUKE UNIVERSITY, The Fuqua School of Business

Master of Science in Quantitative Management: Business Analytics, Finance Domain Coursework: Applied Statistics, Machine Learning, Artificial Intelligence, Data Mining, AWS for Analytics Winner, Winter Data Competition; Chinese Student Network Lead, orchestrating community of 730+ Durham, NC

May 2025

XI'AN INTERNATIONAL STUDIES UNIVERSITY

Bachelor of Economics, Finance

Distinguished Graduate, Meritorious Winner of the America Modeling Competition Shaanxi, China

Jul 2022

EXPERIENCE

SHAANXI QIRUI CO., LTD Shaanxi, China

Data Analyst Jan 2022 – June 2024

• Scraped 100K+ fashion trend records from e-commerce platform Alibaba (Taobao, 1688) using Python (BeautifulSoup, Selenium) to support pricing analysis and new product design; stored cleaned data in AWS RDS using psycopg2 with read replicas, IAM roles, and automated backups, enabling concurrent SQL queries and scalable access for modeling workflows.

• Engineered an Airflow-based ETL pipeline using Python (panda, sqlalchemy) to automate raw data ingestion, feature transformation, and PostgreSQL upload; implemented daily DAGs with retry logic, SlackOperator alerts, and failure logging, reducing reporting latency by 15% and manual refresh burden by 30%.

• Built an XGBoost model to predict supplier payment reliability based on 20K+ historical transaction records across 300+ partners; used SQL (LAG, CTE) to generate behavior flags, late-payment ratios, and rolling 90-day averages; performed multicollinearity checks (VIF < 5) and optimized model via GridSearchCV, achieving 89% test accuracy and reducing delayed order incidents by 20%.

• Built a Random Forest credit scoring model under the PD–LGD framework to assess supplier default risk; combined structured payment records with customer feedback, engineered behavioral and text features, and selected top 15 predictors via feature importance. The model (AUC = 0.91) was deployed in Tableau, reducing overdue receivables by 17% within 3 months.

• Delivered 100+ Tableau and Power BI dashboards for operations, sourcing, and finance teams; applied advanced DAX functions

(e.g., CALCULATE, RANKX) and custom measures to visualize contract risk tiers, aging reports, and fulfillment trends. ByteDance CO., LTD Beijing, China

Data Analysis Intern Jun 2021 – Oct 2021

• Independently cleaned and restructured 67M+ CRM records during advertising platform reorganization; extracted customer IDs, order features, and industry categories, and optimized SQL queries to accelerate access across teams—cutting data response time by ~40% and supporting 3 departments during system transitions under the software development life cycle framework.

• Built 20+ interactive Power BI dashboards to track ad campaign KPIs (e.g., impressions, CTR, ROI), replacing static reporting

(PPT, ThinkCell); authored SOP documents to guide non-technical users in data retrieval and analysis, saving 80% reporting time and enhancing cross-functional communication within Agile delivery cycles.

• Analyzed performance of ad campaigns for key clients (e.g., Shein, Shopee) using A/B testing and various ad formats (e.g., feed ads, splash screens, live links); delivered creative and timing optimization recommendations, boosting client ROI by 1.42x and CTR(Click-Through Rate) by 18%, increasing campaign renewals. SELECTED PROJECTS

Optimization: Healthy & Sustainable Higher Education System (Python, Power BI). Link First Prize of American Mathematical Modeling Competition; Winner, Duke Winter Data Competition Built a data-driven scoring system to evaluate higher education sustainability across 6 countries using 10-year data and 25 indicators; applied AHP and PCA for index construction and dimensionality reduction and used K-Means and Grey Correlation Analysis for performance clustering and policy simulation. Raised Australia’s projected score from 71.3 to 89.9 and proposed an 11-year policy roadmap.

Performance Analytics: Drivers of Venture Capital Fund Success (Python, Tableau, XGBoost). Link Duke Capstone Project

Analyzed 822 venture capital funds across 399 firms by conducting EDA, feature engineering, and dimensionality reduction (PCA); segmented fund archetypes via K-Means and identified structural clusters. Trained an XGBoost regression model (R = 0.96, RMSE

= 2.86) to predict Net IRR based on macro timing, capital deployment pacing, and firm experience, with SHAP analysis revealing macro factors as dominant performance drivers.

TECHNICAL CAPABILITIES

Languages & Tools: Python, R, SQL, Advanced Excel, Tableau, Power BI; Statistical Methods: Regression, A/B Testing, Hypothesis Testing; Certifications: CFA Level II Candidate; Cloud: Basic knowledge of AWS (S3, Lambda, Glue, Redshift), GCP

Contact this candidate