Open to opportunities

Yuexin Zou

@yuexinzou

Data Science undergraduate building ETL pipelines and predictive models, turning complex data into actionable business insights.

United States

Message

What I'm looking for

I’m looking for an entry-level Data Science role where I can build ETL and dashboards, develop predictive models, and collaborate cross-functionally—ideally with meaningful work in trustworthy AI, fairness, and data-driven decision support.

I’m a Data Science undergraduate at the University of North Carolina at Chapel Hill, with hands-on experience in data analytics, business intelligence, and quantitative modeling. I focus on building reliable data pipelines and translating complex datasets into insights teams can act on.

In my internship at Mizuho Orthopedic Systems, Inc., I built a Python-based ETL pipeline integrating datasets from 59 account managers across 7,000 hospitals, cutting manual reconciliation by 40+ hours per month. I also improved reporting speed by 90% by delivering dashboards that surface KPI trends and revenue analytics for senior management. Earlier, as a Business Analyst Intern at BoNiu PE Fund Management, I developed and optimized a credit risk prediction model using XGBoost and LightGBM, improving default prediction accuracy by 9 percentage points through iterative optimization and stakeholder reporting.

Beyond industry work, I’m growing my research skills as an Undergraduate Researcher at UNC’s Society-Centered AI Lab (SAIL), where I’m studying “Fairness Under Scarcity” for LLM agents and mechanical model approaches to homeless shelter allocation. I’ve developed an LLM simulation framework combining chain-of-thought prompting with stochastic Markov modeling and run 300+ experiments to compare fairness outcomes under resource constraints. I also collaborate on uncertainty-aware out-of-distribution materials prediction with a PyTorch GNN pipeline and deep evidential regression, and I’ve built a GAN-based synthetic clinical data generator plus explored LLM-based clinical text generation for data scarcity.

Experience

Work history, roles, and key accomplishments

Current

Undergraduate Researcher

Current

Society-Centered AI Lab (SAIL), UNC

Aug 2025 - Present (10 months)

Developed an LLM simulation framework combining chain-of-thought prompting with stochastic Markov modeling to evaluate fairness in homeless shelter allocation. Ran 300+ experiments comparing LLM agents vs. stochastic optimization models and performed ablations across GPT-4o, Gemini, and Claude Sonnet to assess fairness outcomes.

Chain Of Thought Stochastic Markov Modeling Simulation Experimental Design GPT 4o Gemini

Current

MatUQ Uncertainty Research

Current

MatUQ

May 2025 - Present (1 year 1 month)

Optimized a PyTorch-based GNN inference pipeline using CUDA-accelerated Monte Carlo dropout and deep evidential regression across 6 materials-property benchmarks (140K+ samples). Achieved a 13.6% MAE reduction on sparse-data regimes via hybrid uncertainty quantification.

PyTorch GNN Inference CUDA Deep Evidential Regression Uncertainty Quantification MAE Optimization

Data Analyst Intern

Mizuho Orthopedic Systems, Inc.

Jun 2025 - Aug 2025 (2 months)

Built a Python-based ETL pipeline integrating datasets from 59 account managers across 7,000 hospitals, cutting manual reconciliation by 40+ hours per month. Delivered BI dashboards that accelerated reporting speed by 90% and surfaced KPI trends/revenue analytics for senior management.

Python ETL Data Reconciliation SAP Data Integrity KPIs

CAREGEN Generative AI Project

CAREGEN

Oct 2024 - Mar 2025 (5 months)

Built a GAN-based synthetic clinical data generator for ~3K healthcare datasets with limited training samples. Explored LLM-based clinical text generation to address data scarcity for global healthcare AI use cases.

Synthetic Data Generation LLM Generative AI Data Scarcity Healthcare Data Machine Learning

Business Analyst Intern

BoNiu PE Fund Management

Jun 2024 - Aug 2024 (2 months)

Built and optimized a credit risk prediction model using XGBoost and LightGBM on three years of Wind China market data, applying feature engineering and competitive analysis. Improved default prediction accuracy by 9 percentage points through iterative optimization and stakeholder reporting.

XGBoost LightGBM Feature Engineering predictive modeling Machine Learning Competitive Analysis Credit Risk Modeling

Education

Degrees, certifications, and relevant coursework

University of North Carolina at Chapel Hill

Bachelor of Science, Data Science

2023 - 2026

Grade: GPA: 3.61/4.00

Activities and societies: Dean’s List; Women in Economics Leadership (led 4 mentorship events supporting 200+ freshmen); Teaching Assistant (designed homework, graded, led review sessions and office hours for 300+ students across 3 data courses).

B.S. in Data Science at UNC Chapel Hill (Aug 2023–May 2026) with minors in Statistics and Economics; GPA 3.61/4.00.