Chen Zhao

About Chen Zhao

Quantitative analyst with over 1.5 years of experience developing and implementing advanced risk and marketing models for the finance industry. I specialize in leveraging machine learning, survival analysis, and NLP to drive customer acquisition, assess credit risk, and mitigate operational risk.My approach is built on translating complex data into actionable, high-impact business solutions that optimize marketing funnels and strengthen risk frameworks.Core Competencies & Applications:Marketing & Acquisition Modeling: Building machine learning models (LightGBM, XGBoost) to predict customer conversion, optimize marketing funnel efficiency, and increase loan application approval rates.Behavioral Modeling: Engineering tree-based survival analysis models to forecast customer behavior, such as time-to-next-loan and retention, informing proactive marketing and liquidity strategies.Credit Risk Modeling: Developing, validating, and monitoring core credit risk models (PD, LGD, EAD) using both machine learning and traditional statistical methods (e.g, Cox Proportional Hazards).NLP & Operational Risk: Implementing NLP models to analyze customer communications (e.g, post-collections call data) to predict complaint likelihood and reduce operational risk.Technical Toolkit:Languages: Python (Pandas, Scikit-learn, LightGBM, NumPy), SQLTechniques: Machine Learning (Gradient Boosting, Random Forest), Survival Analysis (Cox, Kaplan-Meier, Tree-based), NLP, Time Series Analysis, Risk EconometricsEducation:My professional experience is built on a strong academic foundation, including an M.S. in Mathematics in Finance from NYU\'s Courant Institute of Mathematical Sciences and a B.S. from the University of Pittsburgh.

Experience

1. Quantitative Researcher

   Xftech

   Sep 2025 — Present

   Developed and deployed LightGBM models to establish a marketing-risk alignment framework; implemented a pre-scoring engine to filter high-rejection leads, mitigating cross-departmental operational friction and achieving a 20% boost in qualified conversion rates by minimizing early-stage funnel loss.2. Fine-tuned BERT models on noisy post-collections call data using an LLM-based semantic denoising pipeline and hierarchical Multi-head Attention; transitioned to a Pairwise LTR framework, resulting in an 80% reduction in customer complaints, a 25% relative improvement in KS, and a 10x Lift in top-tier bins.3. Engineered Uplift models utilizing Label Transformation techniques to estimate CATE for interest rate sensitivity modeling, evaluating the incremental impact of personalized rate discounts on loan drawdown rates. Resolved Control Group Pollution via Inverse Probability Weighting (IPW) to optimize dynamic pricing strategies; successfully isolated rate-sensitive cohorts, driving a 50% drawdown rate improvement in the top 1% targeted segment and a 25% improvement in the top 20%.

Education

• University of Pittsburgh

  Bachelor of Engineering - BE, Materials Engineering

  2018 — 2022

• Georgia Institute of Technology

  Master's degree, OMSCS

• Sichuan University

  Bachelor of Science - BS, Materials Engineering

  2018 — 2020

• New York University

  Master's degree, Financial Mathematics

  2022 — 2023