J同学伦敦大学学院科学和数据密集型计算硕士

学员背景

• 原学校：牛津大学
• 原专业：社会数据科学
• 背景提升：如何提升背景？
• 主要经历：
  2019-07 - 2019-09
  Bohong Capital (private equity), Shanghai, China Researcher (intern)
  •Multifactor trading strategy/Quant/python/MySql
  •Developed a new strategy to enable the current trading system to switch risk preference based on trading history and market performance to maximize Sharpe Ratio.
  •Rebuilt the data structure and inquiry process of current financial database using MySQL with two senior engineers to improve inquiry speed.
  •Developed a unified API including data-feeding, post analysis, back test, … to provide a much powerful platform for other researchers.
  
  2018-08 - 2019-02
  Bohong Capital (private equity), Shanghai, China Researcher (intern)
  •Multifactor trading strategy/Quant/python/MATLAB
  •Applied the principle of machine learning on Fama and French Three Factor Model to produce factor with high IC IR. Classical methods including regression, SVM, boosted tree were also used as benchmark.
  •Analyzed the characteristics of trading behavior in local market and optimized the trading time window and trading frequency to maximize portfolio profit and reduce max draw down.
  •Designed and built a back-test tool for above trading strategy which will be compatible with most of the future feature including risk management, portfolio combination, …
  •The portfolio is about $2 million cash flow annually.
  
  2017-10 - 2017-12
  Oxford Instrument, Oxford, UK Researcher (intern)
  •Physics/bash/python/MATLAB
  •Measured the electrical resistance of Cobalt under a variety of temperature and pressure condition computationally and experimentally.
  •Simulated the critical surface of resistance of metal with different compounds using Opera and MATLAB.
  •Verified the simulation result experimentally cooperated with Amalia Coldea, professor of Oxford.
  
  2017-07 - 2017-09
  INTEGEO.LLC, Houston, USA Researcher (intern)
  •Geological prospecting data post analysis/Physics/python/MATLAB
  •Cleaned source seismic data and filled missing data using OLS.
  •Recorded the performance of machining learning method, including supervised learning (NN, CNN) and unsupervised learning (k-means clustering), on recognizing different geological configurations using classical PDE method as benchmark.
  •Improved the outcome result/figure of PDE method, including local resolution improvement, geological fault recognition and repair.