Developed the company’s flagship AI-driven fund focused on REITs, achieving approximately 20% annual returns. The fund is built around three core components:• Fundamental Analysis: Sourced and finalized data vendors, cleaned annual report data, and constructed Fama-French factors. Applied advanced Fama-French models to optimize REIT selection.• Technical Analysis: Processed monthly trading data including millions of data rows and developed new features based on academic papers and research. The primary model identifies high-return REIT categories, while sub-models, using machine learning algorithm to select individual tickers within those categories.• Sentiment Analysis: Scraped news data, calculated correlations between REITs and news articles using Glove embeddings, and analyzed sentiment with TextBlob and vader. Visualized results through sentiment graphs using Matplotlib.

Utilized HeidiSQL to obtain, search and clean data thoroughly through MySQL. Transfer data to Python and disposal using packages (NumPy, Pandas, Xarray). Improved quantitative ETF strategy by increasing Sharpe ratio and annual yield. Selected indices by adjusting parameters in machine learning including random forest and neural network. Created signal to help support buying and selling decisions which reached 24% annual yield.

Compiled research reports of securities companies, produced minute and daily frequency. Conducted evaluation and back testing, such as Hide Track factor, SJVR factor, etc. and can reach 7% IC value and increased alpha return. According to future trade rule, transformed stock factors to future factors.

Researched quantification strategies and completed back test on quantification platform, such as alpha strategy, momentum reversal strategy, etc. Besides, rewrote alpha lens packages in Python for factor testing, prepared codes and theoretical knowledge related to financial engineering.