Education

GEORGIA INSTITUTE OF TECHNOLOGY, Atlanta, GA

Master of Science in Quantitative and Computational Finance, Scheller College of Business, Aug 2019 – Dec 2020

• Key Coursework: Finance and Investments, Stochastic Processes in Finance I, Derivative Securities, Fixed Income Securities, Design and Implementation of Systems in Computational Finance, Numerical Methods in Finance, Financial Optimization, Financial Data Analysis, Computational Data Analysis

NATIONAL UNIVERSITY OF SINGAPORE, Singapore

Doctor of Philosophy in Mathematical Finance, Department of Mathematics, Jan 2014 – May 2018

• Math PhD Qualifying Exam: 80/80 (average score among 15 candidates: 35/80)
• Journal Publications (alphabetical author order): Opaque bank assets and optimal equity capital (with Min Dai and Jussi Keppo) on Journal of Economic Dynamics and Control, (100) 369–394.

BEIJING NORMAL UNIVERSITY, China

Bachelor of Science in Applied Mathematics, Department of Mathematics, Sep 2009 – July 2013

• GPA: 94.53/100 (Ranking: 1/101, Top 1%)
• Honors: Outstanding Bachelor’s Thesis (3%); National Scholarship (1%), China.

Working Experience

United Overseas Bank Limited (UOB) , Singapore

Senior Officer, Risk Analytics of Market Making and Group Global Markets (Python, VBA), Jan 2019 – Aug 2019

UOB is a leading company in Asia with a global network of more than 500 braches and offices in 19 countries and territories in Asia Pacific, Europe and North America. UOB provides commercial and corporate banking services, personal financial services, and insurance services.

• Validated quantitatively Market Risks among different market groups include equity products and derivatives, commodity, fixed income, FX derivatives, credit derivatives, and interest rate derivatives.
• Performed quantitative research on market risk. Built multilayer perceptron (MLP) neural networks to fast approximate the of price early excisable products (see e.g. American option) and to fast build and extract volatility surface.
• Implemented machine learning and Fintech application in Credit Risks. Built Gradient Boosting Classification model to beat the classical FICO scorecard model. Higher IV, KS, and AOC are achieved.
• Performed analysis on ML model sensitivity, stability, and assessing trust in modelling financial products including model selection, hyperparameter auto-tuning via Bayesian optimizations, and model explanation and trust via LIME and SHAP.

National University of Singapore, Singapore

Postdoc, School of Computing and Institute of Operations Research and Analytics (Python) , July 2018 – Dec 2018

• Research on Fintech including smart contract with BlockChain, probability of default (POD), price prediction, news detection, and stochastic control by reinforcement learning.
• Built new dynamic model for banks that face liquidation risks and optimize simultaneously dividends, recapitalizations, and asset fire sales with adjustment costs over time. Cost shocks were involved in the structural estimation.

Skills

Programming: Python, C++, R, SAS, VBA, Matlab, Stata

Preprint

• Opaque bank assets and optimal equity capital (with Min Dai and Jussi Keppo) on Journal of Economic Dynamics and Control, (100) 369–394.

Abstract: Banks’ assets are opaque, and therefore, we model their true accounting asset values as partially observed variables. We derive a stochastic control model to optimize banks’ dividend and recapitalization policies in this situation, and calibrate that to a sample of U.S. banks. By the calibrated model, the noise in reported accounting asset values hides about one-third of the true asset return volatility and raises the banks’ market equity value by 7.8%. Particularly, those banks with a high level of loan loss provisions, nonperforming assets, and real estate loans, and with a low volatility of reported total assets have noisy accounting asset values. Because of the substantial shock on the true asset values, the banks’ assets were more opaque during the recent financial crisis.

• Optimal Portfolio Choice for Early Retirement with Cointegration between the Stock and Labor Markets (with M. Dai and Seyoung Park)

Abstract: Existing literature concludes that there exists a certain threshold of wealth for retirement. We show that this conclusion counts crucially on the assumption of short-term income risk. We derive early retirement strategy with long-run income risk characterized by cointegration between the stock and labor markets. We find that the derived wealth threshold for retirement varies with levels of long-run income risk. In particular, earlier retirement with a lower wealth than scheduled without long-run income risk could be optimal as labor income is expected to decrease than as it is expected to increase in the long term.

• Asymptotics for Minimizing Lifetime Ruin Probability with Small Transaction Costs (with Xinfu Chen, Min Dai, and Bin Li)

Abstract: In this paper we optimally minimize the lifetime ruin probability for an investor when she faces small transaction costs in stock trading. We perform technical asymptotic analysis of the optimal buy and sell boundary in the presence of small transaction cost. Specially, when the investor's position is close to the ruin boundary, she tends to trade more conservatively and hold less stocks by enlarging the sell region and shrinking the buy region dramatically.

• Optimal Portfolio Choice with Cointegration between Stock and House Markets (with Yingshan Chen, Min Dai, and Hong Liu)

Abstract: Even though stock and housing markets have low contemporaneous correlations, they are cointegrated. We show that in the presence of cointegration, households significantly increase housing expenditure, reduce stock investment, and may choose not to participate in the stock market at all. In addition, for a given level of wealth, the critical level of the participation cost above which households never participate in the stock market is much lower. Our model can thus potentially help explain both the puzzle of stock market non- or limited-participation and the puzzle of the highly negative correlation between stock and housing investment.

• The Optimality of Simple Banking Regulations (with Jussi Keppo and CheLin Su)

Abstract: We derive a new dynamic model for banks that optimize simultaneously dividends, recapitalizations, and asset fire sales. Our model can explain the substantial drops in capital ratio, bank value, and dividend payments, as well as the promotions in recapitalizations and asset fire sales during the financial crisis of 2007--2008. Further, we show that compared to refinancing, dividend restriction accelerates the recovery of banking system faster when banks' equity capital ratios fall low. It is also helpful for government to provide guaranteed fund to banks that are closer to the default point during recession.

• Good or Bad News Principle on Corporate Investment (with Wei Jiang and Shuajie Qian)

Abstract: This paper examines optimal corporate investment policies by quantifying the impact of time-to-build and demand uncertainty on investment given the existing capital stock. Investment is characterized by construction lag and costly reversibility when a firm can purchase capital at a given price and sell capital at a lower price. We attempt to propose a unified theory about how investment frictions, including investment lags and costly reversibility, together with demand uncertainty, affect a firm's investment strategies. Our model encompasses the case of irreversible investment as well as the standard neoclassical case of costlessly reversible investment.

• Costly Bank Recapitalization with Execution Delay

Abstract: How should shareholders make dividend and recapitalization policies for a bank when there exists execution delay and fix costs on recapitalizations? What is the optimal time and magnitude of recapitalization if the magnitude of recapitalization is pre-determined at the time of recapitalization-making, instead of at the delayed time of execution? How much impact will the execution delay have on bank policies? To answer these questions, we develop a continuous time model based on the work of Peura and Keppo [The Journal of Business 79(4), 2163-2201, 2006] (henceforth PK) and reformulate the path-dependent problem to a path-independent problem by introducing an additional state variable to record the recapitalization amount. An iteratively numerical algorithm is provided to solve the problem and can be widely applied to other financial models with execution delay or decision lag. Compared to PK, the information loss due to execution delay induces more conservative and rational policies to keep higher level of capital ratio and thus, lowers the bank value.

• Taxable Stock Trading with Deep Reinforcement Learning

Abstract: In this paper, we propose stock trading based on the average tax basis. Recall that when selling stocks, capital gain should be taxed while capital loss can earn certain tax rebate. We learn the optimal trading strategies with and without considering taxes by reinforcement learning. The result shows that tax ignorance could induce more than 62% loss on the average portfolio returns, implying that taxes should be embedded in the environment of continuous stock trading on AI platforms.

Conference Presentations

• (Invited) Oral Presentation at the 2nd Paris-Asia Conference in Quantitative Finance (Suzhou, China), May 2017.
• (Invited) Oral Presentation at the 9th World Conference of the Bachelier Finance Society (New York City), July 2016.
• (Invited) Oral Presentation at Berlin-Princeton-Singapore Workshop on Quantitive Finance (Princeton University), June 2016.
• (Invited) Oral Presentation at the Interdisciplinary Approaches to Financial Stability Conference (University of Michigan), October 2015.