Yunxiao Xiang
YUNXIAO XIANG
------------ ? ------------ ? ------------/in/yunxiaoxiang
EDUCATION
New York University, The Courant Institute of Mathematical Sciences New York, NY
M.S. in Mathematics in Finance; GPA: 3.8/4.0 Dec. 2020
• Coursework: martingales, Monte Carlo, local volatility, SVI, Brownian motion, Black-Scholes, VaR, Greeks, Itô lemma, GARCH, cross-validation, LSA, LDA, random forest, Kalman filter, K-NN, boosting, Gaussian kernel, reinforcement learning, neural network, B-spline, Bayesian inference, SVM, clustering
University of California, San Diego La Jolla, CA B.S. in Applied Mathematics; B.A. in Economics; GPA: 3.8/4.0 Jun. 2019
• Coursework: Markowitz model, CAPM, arbitrage pricing theory, factor model, hypothesis test, ODE, bootstrap, MLE, CLT, SVD, PCA, regression, ACF, ARIMA model, backtesting, heat equation, GA
SKILLS
Programming Languages: Python (6 years), Java (5 years); R, Excel, MATLAB (2 years); SQL (1 year)
Tools/Technologies: AWS Services, LaTeX, Git, Bloomberg, JUnit, OOP, test-driven development (TDD)
Operating Systems: Linux, macOS, Windows IDEs: Jupyter, Sublime, Eclipse, PyCharm, RStudio, MySQL
EXPERIENCE
RavenPack New York, NY
Summer Research Project Leader (Mentors: Ricard Matas, Peter Hafez) Jul. 2020 – Sep. 2020
• Filtered for novel events based on sentiment score; visualized distance between events and analyst ratings
• Leveraged Bayesian approach to compute P (analyst rating change | event X happened in Y days) for each (X, Y, entity); checked event volume, probability distributions and significant ratios for subset selection
• Implemented XGBoost to forecast analyst rating events; tackled imbalanced labels by oversampling
• Constructed signals to build event-driven portfolio; evaluated prediction power and portfolio metrics
Axiomquant Investment Management, LLC HQ: Beijing, CN
Quantitative Research Intern (Remote in New York) Jul. 2020 – Sep. 2020
• Processed 5 years’ auction, close, market data to extract 132 intraday, cross-date, cross-stock features
• Leveraged LRU cache to optimize repetitive function call, multiprocessing to parallelize computation
• Built OLS, Ridge, and elastic net to predict future returns; selected significant features by out-of-sample liquidity-weighted correlation, rolling cross-validation, Sharpe and PnL of prediction-based portfolio
• Backtested daily rebalanced portfolio on test set; achieved correlation of 0.087 and excess Sharpe of 1.57
Ubiquant Investment Co., Ltd. HQ: Beijing, CN
Data Analyst Intern (Remote in New York) Apr. 2020 – Jul. 2020
• Implemented Almgren’s impact model to estimate implicit cost of trades size up to 10% of market volume
• Processed TAQ data to efficiently generate model inputs – volume time, execution details, volatility, etc.
• Leveraged non-linear Gauss-Newton optimization and regression to fit impact coefficients and exponents
• Incorporated trading impact in backtesting strategy to compute more realistic Sharpe (from 4.38 to 3.53)
PROJECTS
Identifying Similar Articles – Latent Sentiment Analysis in Python
• Tokenized Reuters Article data with TF-IDF; applied truncated SVD to reduce dimensionality to 400
• Leveraged K-NN classifier to assign corpus to closest category of documents based on cosine similarity
Deal Probability of Russian Commodities – NLP in Python and Multivariate Regression in R
• Leveraged NLP, GPT, and image processing to extract numerical variables from descriptions and images
• Built logistic regression after cross-validation and subset selection to model skewed deal probabilities
Reducing Time Complexity of TSP – Genetic Algorithm and Multithreading in Java
• Implemented shuffle, crossover, tournament select, mutate; reduced search space by factor of 10,000
• Coded synchronized multithreading methods to increase speed of fitness computation by 4 tim¬es
YUNXIAO XIANG
------------ ? ------------ ? ------------/in/yunxiaoxiang
EDUCATION
New York University, The Courant Institute of Mathematical Sciences New York, NY
M.S. in Mathematics in Finance; GPA: 3.8/4.0 Dec. 2020
• Coursework: martingales, Monte Carlo, local volatility, SVI, Brownian motion, Black-Scholes, VaR, Greeks, Itô lemma, GARCH, cross-validation, LSA, LDA, random forest, Kalman filter, K-NN, boosting, Gaussian kernel, reinforcement learning, neural network, B-spline, Bayesian inference, SVM, clustering
University of California, San Diego La Jolla, CA B.S. in Applied Mathematics; B.A. in Economics; GPA: 3.8/4.0 Jun. 2019
• Coursework: Markowitz model, CAPM, arbitrage pricing theory, factor model, hypothesis test, ODE, bootstrap, MLE, CLT, SVD, PCA, regression, ACF, ARIMA model, backtesting, heat equation, GA
SKILLS
Programming Languages: Python (6 years), Java (5 years); R, Excel, MATLAB (2 years); SQL (1 year)
Tools/Technologies: AWS Services, LaTeX, Git, Bloomberg, JUnit, OOP, test-driven development (TDD)
Operating Systems: Linux, macOS, Windows IDEs: Jupyter, Sublime, Eclipse, PyCharm, RStudio, MySQL
EXPERIENCE
RavenPack New York, NY
Summer Research Project Leader (Mentors: Ricard Matas, Peter Hafez) Jul. 2020 – Sep. 2020
• Filtered for novel events based on sentiment score; visualized distance between events and analyst ratings
• Leveraged Bayesian approach to compute P (analyst rating change | event X happened in Y days) for each (X, Y, entity); checked event volume, probability distributions and significant ratios for subset selection
• Implemented XGBoost to forecast analyst rating events; tackled imbalanced labels by oversampling
• Constructed signals to build event-driven portfolio; evaluated prediction power and portfolio metrics
Axiomquant Investment Management, LLC HQ: Beijing, CN
Quantitative Research Intern (Remote in New York) Jul. 2020 – Sep. 2020
• Processed 5 years’ auction, close, market data to extract 132 intraday, cross-date, cross-stock features
• Leveraged LRU cache to optimize repetitive function call, multiprocessing to parallelize computation
• Built OLS, Ridge, and elastic net to predict future returns; selected significant features by out-of-sample liquidity-weighted correlation, rolling cross-validation, Sharpe and PnL of prediction-based portfolio
• Backtested daily rebalanced portfolio on test set; achieved correlation of 0.087 and excess Sharpe of 1.57
Ubiquant Investment Co., Ltd. HQ: Beijing, CN
Data Analyst Intern (Remote in New York) Apr. 2020 – Jul. 2020
• Implemented Almgren’s impact model to estimate implicit cost of trades size up to 10% of market volume
• Processed TAQ data to efficiently generate model inputs – volume time, execution details, volatility, etc.
• Leveraged non-linear Gauss-Newton optimization and regression to fit impact coefficients and exponents
• Incorporated trading impact in backtesting strategy to compute more realistic Sharpe (from 4.38 to 3.53)
PROJECTS
Identifying Similar Articles – Latent Sentiment Analysis in Python
• Tokenized Reuters Article data with TF-IDF; applied truncated SVD to reduce dimensionality to 400
• Leveraged K-NN classifier to assign corpus to closest category of documents based on cosine similarity
Deal Probability of Russian Commodities – NLP in Python and Multivariate Regression in R
• Leveraged NLP, GPT, and image processing to extract numerical variables from descriptions and images
• Built logistic regression after cross-validation and subset selection to model skewed deal probabilities
Reducing Time Complexity of TSP – Genetic Algorithm and Multithreading in Java
• Implemented shuffle, crossover, tournament select, mutate; reduced search space by factor of 10,000
• Coded synchronized multithreading methods to increase speed of fitness computation by 4 tim¬es