Data Scientist - Predictive Modeling & ETL Expert

Zhaoxuan Jin

Dallas,TX

+1-469-***-**** ********.***@*******.*** https://www.linkedin.com/in/zhaoxuan-jin-8a76b6204/ SUMMARY

Data professional with expertise in ETL pipelines, machine learning–based predictive modeling, data automation, and business analytics using Python, SQL, Power BI, and Excel. Proven experience designing and implementing data-driven solutions — from risk underwriting models that improved delinquency detection by 20%, to Power BI dashboards enabling real-time loan servicing insights, and hybrid ML + business-rule forecasting systems automating daily auction bid optimization. Dedicated to developing data-driven products that bridge predictive analytics with real-time operational execution. TECHNICAL SKILLS:

• Programming & Analytics: Python (Pandas, NumPy, scikit-learn, Matplotlib, statsmodels), PySpark(DataFrame API, Spark SQL)

• Data Visualization: Power BI (DAX, data modeling), Excel (Power Query, PivotTables, lookup functions)

• Databases: SQL Server (complex joins, window functions, temp tables, stored procedures) BUSINESS EXPERIENCE

Tekzenit Dec 2023 - Oct 2025

Software Engineer – Data Science Irving, TX

Risk Underwriting Model

• Developed a predictive underwriting model for early-stage delinquency risk, segmenting borrowers into four FICO-based groups to capture differentiated risk behaviors and strengthen portfolio risk assessment.

• Performed comprehensive feature engineering, including IV-based variable selection, monotonic WOE transformation, and multicollinearity reduction. Applied AIC-based stepwise regression to iteratively refine variables and fit the final logistic model, ensuring both interpretability and stability across borrower segments. In parallel, benchmarked Random Forest feature importance

(top 50 features) as an alternative approach, confirming stepwise as the most effective method.

• Benchmarked the new model against the existing production model, delivering a +8 KS uplift and Gini improvement (0.35 0.52)

, and validated robustness on out-of-time datasets to monitor drift and long-term performance; ultimately improved delinquency detection accuracy by 20% and reduced portfolio risk. Loan Servicing Dashboard

• Developed an interactive Power BI dashboard for loan performance monitoring across five key dimensions: Loan Summary, DPD, Roll Rate, Payment Trends, and Call Activity.

• Implemented dynamic filters and visuals to deliver real-time portfolio insights, enabling business leaders to track delinquency trends, monitor repayment behaviors, and make faster, data-driven decisions.

• Built ETL pipelines (DealerSocket IDMS/CRM SQL Server Power BI) and optimized SQL Server stored procedures and temp tables (e.g., roll rate modeling, aging buckets, call analytics) to support accurate KPIs like Delinquency %, Roll Forward Rates, and Payment Performance, enabling better risk segmentation and early intervention. Auto Auction Bidding Optimization

• Developed a vehicle price forecast model and ROI optimization framework integrating BasePrice, 3-week median trend, MMR reference, and competition adjustment to predict next-week bid prices; established a weighted model to balance win-rate improvement and margin protection.

• Built an end-to-end SQL + Python automation pipeline for daily auction preparation — generating ROI-adjusted, model-based bid recommendations for 500+ vehicles per day, performing duplication and ownership filtering, executing automated real-time data upload with data integrity checks, and integrating with the bidding system within strict daily time windows.

• Designed and delivered Excel reports (Purchased, Missed, Removed, Owned Units), enabling bid review, performance tracking, and inventory & budget planning; reduced manual consolidation by 75% and improved bidding accuracy by 25%. EDUCATION

The University of Texas at Dallas Dec 2023

M.S., Information Technology and Management GPA: 3.809 Harbin Normal University

B.A., Teaching Chinese as a second language Jun 2014

Contact this candidate