- Designed new energy-based machine learning model for time series prediction, which improved the old baseline model by 20% in terms of negative log-likelihood- Built LSTM model with attention in J language to enhance return on a $300M algorithmic fund- Vectorized existing neural network models by J language and shortened training time by 50% of an algorithm running on Kdb and Jd database- Implemented variations of LSTM, RNN, VAR models in J and Python to facilitate algo research- Developed a web scrapper in Ruby to collect stock price data and calculate momentum indicators

- Implemented and backtested R-Breaker momentum/reversal trading strategy in R and excel- Built and tested an Empirical Mode Decomposition (EMD) indicator algorithm in R, which had directional prediction accuracy over 70% on CSI 300 stock index over 2014-2017- Researched topics in Generalized Extreme Value distribution and Bayesian inference to facilitate development of a new commodity futures trading strategy

- Drafted prospectus and due diligence reports for a CNY 6.4B bond issuance deal- Presented on bond issuance details in pre-offering roadshows for the deal above- Wrote research reports on bond default cases and industry credit analysis (steel and real estate)