Xiangpeng Zhao

***W **st St, New York, NY *****

+1-917-***-**** ac771u@r.postjobfree.com www.linkedin.com/in/xiangpengzhao EDUCATION

Columbia University in the City of New York, New York, NY Aug 2017-Dec 2018(expected) M.S. in Operations Research, Financial and Managerial Applications Concentration, GPA: 3.73 Course Assistant: IEOR 4720 Deep Learning

Courses: Optimization, Stochastic Models, Machine Learning, Model Based Trading, Deep Learning, Simulation Programming Languages: Python, R, C++, Tensor Flow, HTML, SQL, LabVIEW Shanghai Jiao Tong University, Shanghai, CN Sep 2013- Jun 2017 B.S. in Thermal Energy and Power Engineering, GPA: 3.61 Courses: Probability & Statistics, Object Oriented Programming in C++, Data Structure & Algorithm, Mathematics of Physics PROFESSIONAL EXPERIENCES Transcendence Capital Management Quantitative Researcher Intern New York, Jun 2018 - Aug 2018

• Alpha Research: Constructed trend-following and swing trading strategies on bitcoin based on the research for over 20 popular indicators and strategies. The best improved momentum strategy achieved around 500% annualized return, a Sharpe ratio of 7.74 in the back test, and 16% return in one month in the real market.

• Genetic Algorithm: Developed a parameter tuning tool in python by implementing genetic algorithm which significantly saved training time from one week to 12 hours.

• NLP Data Prep: Applied deep learning models in the NLP project. Cleaned and resampled cryptocurrencies’ tick data. Syrah Capital Quantitative Analyst Intern New York, Apr 2018 - Jul 2018

• Neural Network Tool Development: Built a handy and multifunctional NN system in python to facilitate the process of finding daily arbitrage opportunity of stocks and ETFs.

• Multiclass classification: Classified training data’s daily return into up, down or stationary so the trading system could only make transactions when it’s confident. Importance of three classes can be easily adjusted in training based on trader’s risk preference and trading frequency.

• Feature Selection: Discovered and selected technical indicators such as RSI, CCI as the input for the neural network. Soochow Securities Research Intern Shanghai, Jun 2017 - Aug 2017

• Statistical Analysis: Analyzed a portfolio of stocks in the environmental industry. Implemented Jarque-Bera test in terms of kurtosis and skewness. Investigated serial and cross correlations. Estimated VaR.

• Equity Research: Supported evaluations toward environmental industry by analyzing data, policies and wrote reports. RESEARCH EXPERIENCES

Event-driven Stock Price Prediction New York, Jan 2018 – May 2018

• Image Classification: Replicated and extended Hinton’s dropout paper. Applied image prepossessing and CNN to MNIST, SVHN, CIFAR-10, then prevented overfitting by using dropout or L2 normalization. Tried to beat the paper values by including more modern techniques, such as using Adam optimizer, implementing dropout CNN and batch-norm CNN.

• NLP and Prediction Model: Trained the event-embedding with FastText and Reverb and treated financial news text and event-embedding as input to learn the predictive power based on CNN, LSTM and GRU and predict next day’s price movement direction. The best accuracy of the research is 70.2% achieved by GRU, which beats 66.93% of the paper. Reinforcement Learning Trading System New York, Jan 2018 – May 2018

• Policy Gradient: Constructed a trading strategy on stock market using policy gradient, a classic reinforcement learning algorithm. Used a multilayer neural network which generated probability of actions based on states, and iteratively adjust itself in the scale of the reward by taking those actions.

• Execution Strategy: Applied VWAP trading strategy to confirm the feasibility of transactions generated above. Quantitative Trading Strategy New York, Sep 2017-Dec 2017

• Alpha Research: Developed a python-based trend following strategy on the one-day timeframe. Generated and filtered trading signals by utilizing Fibonacci moving average, fractal theory, oscillator, accelerator oscillator and MACD filter.

• Algorithm Trading: Live trading on Interactive Brokers, including multiple exit strategies. The Sharpe Ratio was 1.26 in the real market.

• Multithreaded project (C Implemented the optimal trade execution dynamic program under the generalized trading impact model which significantly reduced trading cost.