• Studied quantitative trading of cryptocurrencies, implemented Stochastic Oscillator strategy, built backtesting framework and completed parameter optimization, backtesting 3-year cumulative return of over 110%• Built a crawler using Python and Selenium tools to obtain Ethereum transaction data and continuously maintain and update the company's relevant on-chain database• Obtained encrypted Uniswap transaction logs and smart contract ABI from Etherscan and decode them using Python (eth_utils, web3) and Javascript• Built a dashboard product based on the Django framework for on-chain risk analysis on blockchain, with risk scoring and risk path visualization for cryptocurrency transaction blocks

• Adopted the Engel-Granger method to conduct cointegration tests on commodity futures and their respective near- and far-month contract price series, and designed inter-period arbitrage and cross-species arbitrage strategies based on the grid trading method, involving products such as copper, aluminum, zinc, rebar, plastics, soybean meal, coking coal, etc.• Optimized strategy parameters such as grid spacing, quantity and price pivot using BP algorithm to control the maximum retracement of strategy back-testing on multiple species to within 20%

• Synthesize stock selection factors based on convolutional neural networks, automatically combine factor data from multiple time sections into nonlinear features, and test new factors in the A-share market• Extracted the news sentiment scores of the major constituents of the CSI 300 stock index and examined the correlation and lead-lag relationship between the sentiment scores and stock price movements• Engaged in research on cryptocurrency trading patterns, cleaning cryptocurrency tick data, building backtesting frameworks, optimizing parameters such as pending prices and pending times, and helping to reduce slippage in algorithmic trading

• Built a fundamental factor library for enterprises, tested factor validity according to robust regressive coefficients and information coefficients, selected available factors from the factor library using LightGBM algorithm and Recursive Feature Elimination algorithm, and chose the best portfolio within the stock pool based on cross-sectional factor stratification returns, time series rolling portfolio returns and Sharpe ratios • Applied Extreme Gradient Boosting algorithm, Random Forest model and Support Vector Machine to conduct multi-factor stock selection, constructed dynamic portfolios and time-based trading strategies • Used Python's gplearn package to mine new linear and nonlinear stock selection factors based on genetic planning, utilized factor mutual information and multiple excess returns as adaptive evaluation functions in the algorithm, and performed cross-validation to control for possible over-fitting problems

• Managed customer risk databases using Hive, researched the theoretical basis of the off-site early warning models, designed risk screening models based on credit violation logic, created, linked, and maintained user supervision databases

• Updated daily data on companies issuing dollar bonds, conducted due diligence of offshore dollar bond issuers in order to analyze and evaluate the value of investment in several market segments using Wind and Bloomberg • Acquired data from various websites and databases, selected benchmark and comparable bonds, performed data processing and analysis work to write weekly Chinese dollar bond market reports, and completed credit evaluation studies and contract reviews

• Investigated the possible application of community classification algorithms for financial fraud risk monitoring • Constructed a knowledge map and mined association rules using identity information, affinities, judicial files and communication records of Internet finance participants obtained from internal company database • Applied the modular function and the local heuristic removal algorithm Louvain to perform community segmentation and categorize Internet finance users into potentially risky groups • Visualized knowledge graphs using the graphical database Neo4j and Python back-end