1. Synthesized transaction data from Dune and EigenPhi using SQL to gauge sandwich attack prevalence on CoW Swap; proposed mitigation strategies in reports, securing $2,700 in rewards2. Leveraged Python and Tableau to assess impact of ETH Shapella upgrade on liquidity staked token prices such as rETH and stETH, as well as quantity of related MEV arbitrage, sandwich attacks, and liquidation activities, providing valuable insights for 10,000+ users3. Conducted an in-depth examination of DODOEX, uncovering sandwich attacks contributed to over 60% of total volume, and identified shortcomings in 1inch aggregator's slippage settings, leading to potential solutions to enhance user trust and protection

1. Conducted sentiment analysis of crawled tweet text about certain assets in Random Forest, NRC and Vader classifier respectively, and "compound" value obtained by Vader classifier was considered as sentiment classification result of text.2. Introduced above classification results on basis of autoregressive model, trained model and back-tested it, and obtained R square value of model close to 0.98 and MSE of 13, has good prediction ability.

1. Directed an intern team of four, spearheaded improvement for housing rental policy promotion; designed and implemented a cutting-edge bespoke QR code generator, resulting in 70% surge in customer acquisition and 2 million RMB rental contracts2. Formulated visualization reports with Python to measure promotion effectiveness, analyzed correlation between number of QR code scans and conversion rates, generated insights to inform adjustments for promotion strategies3. Coordinated cross-functional teams, arranged project pilot test end-to-end execution, presented to senior leaderships and received top-notch reviews

1. Managed a team to optimize recruiting resume recognition model accuracy from 91.2% to 94.4%, with Human-in-the-loop strategy2. Conducted extensive data labelling to refine model evaluation indicators, implemented data visualization and analysis of annotation results in Python3. Updated main database with Python NLP packages to detect latest vs. obsolete information similarities, resulting in 6% corrections

1. Handled the main business of the bank, such as capital business, financial products, and financial management.2. Connected the front-end and back-end teams of financial products in time; tracked the feedback of products to improve them.