My background combines statistics and stochastic processes, financial markets, and machine learning. I hold an M.S. in Applied Data Science from USC and a B.S. in Statistics and Data Science from UC Santa Barbara. As a WorldQuant BRAIN Research Consultant, I research systematic equity signals; my independent work also covers derivatives valuation, volatility, market microstructure, and trading strategy evaluation.

I work across the research process: forming an economic hypothesis, translating it into a model, testing it against data, and examining whether the result survives robustness checks, portfolio overlap, and execution costs. The projects below connect that research with Python implementations and documented experiments.

University of Southern California — Viterbi School of Engineering
M.S. in Applied Data Science · January 2026

Graduate coursework in Machine Learning, Data Mining, Database Systems, and Bias in Large Language Models complemented my undergraduate statistical training with predictive modeling, data systems, and model evaluation.

University of California, Santa Barbara
B.S. in Statistics and Data Science

My undergraduate coursework spans three foundations of quantitative finance:

• Stochastic and mathematical modeling: Stochastic Processes I & II, Differential Equations, and Loss Models.
• Statistical inference and empirical research: Regression Analysis, Nonparametric Methods, and Experimental Design.
• Markets and implementation: Derivative Markets, Fixed Income Markets, and Python and Object-Oriented Programming.

This training connects probability and inference with the instruments being modeled, while my research projects apply those foundations to signal selection, pricing, and portfolio decisions.

As a WorldQuant BRAIN Research Consultant, I research equity signals across price and volume, fundamentals, analyst estimates, options, earnings, news, and sentiment. My work includes hypothesis generation, simulation, turnover and robustness analysis, and correlation-aware selection.

In the 2026 International Quant Championship, I ranked 578th of 22,085 Stage 2 teams worldwide and 15th in the United States. I selected 36 submitted alphas from 31,448 simulations, comparing dataset-led exploration, systematic sweeps, paper-inspired ideas, and LLM-assisted research. The final competition ranking used an out-of-sample score; the individual alpha diagnostics in the research case study are in-sample platform results.

My research from April 2026 to the present has also led me to design two systems addressing different parts of alpha discovery:

• Adaptive Alpha Scanner — Where should the next backtest be spent? I designed a sequential experiment combining Wald sequential probability ratio testing, diverse dataset screening, and adaptive field prioritization to allocate a fixed simulation budget. The public implementation demonstrates sampling, stopping, budget control, and resumability, with explicit treatment of dependence and selection bias.
• Alpha Factory — How can research hypotheses be generated and refined systematically? I built an LLM-guided workflow that interprets field semantics, proposes testable hypotheses, and uses measured failures to guide bounded revisions. Deterministic software controls validation, deduplication, budgets, stopping rules, and human review; an experiment ledger records candidate lineage and decisions. The documented run covers 102 ideas and 523 backtests; no candidate has passed the complete acceptance criteria.

Together, these projects extend my signal research into the design of research methods and software: deciding where to search, structuring experiments, and retaining the evidence needed to evaluate each result.

572nd of 18,803 teams overall (top 3.1%) · 10th in China · 225th in the manual track. Across five rounds, I developed trading algorithms and made decisions involving quoting, inventory, options, auctions, and portfolio construction. My competition retrospective reconstructs original submissions against exchange logs and examines fair value, signal persistence, liquidity, and execution costs, distinguishing competition results from subsequent research.

I trade options discretionarily, with a focus on volatility, option pricing, Greeks, position construction, and risk management. This experience informs the questions I explore in derivatives research: how exposures evolve, what drives P&L, and where pricing approximations become unreliable.

A connected valuation framework spanning SOFR curve construction, SABR volatility calibration, option pricing, Greeks, P&L attribution, and portfolio risk. The central case prices an eight-leg SPY option book and compares Greek approximations with full repricing. Second-order terms reduce mean unexplained daily P&L from $348 to $43 in the documented study; stress tests expose where the local approximation breaks down. The repository includes calibration diagnostics, VaR/ES backtests, and model limitations.

Two-Stage Recommendation, Bandits & Experimentation — A recommendation research pipeline combining collaborative-filtering retrieval, LambdaMART ranking, contextual bandits, and offline experimentation. On the documented Yelp evaluation, item-CF retrieval outperformed both the popularity baseline and the two-stage ranker, making model comparison and diagnosis central to the study.

Credit Portfolio Analytics — A consumer-credit case study connecting SQL portfolio analysis, default modeling, risk-return pricing, and customer segmentation. It translates model scores into policy scenarios and an executive dashboard, with AUC 0.767 on the reported evaluation and explicitly stated economic assumptions.

Python · SQL · R
NumPy, pandas, SciPy, statsmodels, scikit-learn, XGBoost, PyTorch, PySpark; SQLAlchemy, SQLite, Pydantic, Git, AWS EC2, and SageMaker.