About Me

EDUCATION

University of Chicago

Master of Science in Statistics

Overall Average GPA: 3.89/4.0

• Relevant Coursework:
Statistical Theory and Methods, Applied Linear Statistical Methods, Generalized Linear Models, Multivariate Statistical Theory, Time Series and Forcasting, Big Data, Machine Learning, Deep Learning, Natural Language Process, Bayesian Modeling, Causal Inference, Statistical Inference

National Taiwan University (NTU)

Bachelor of Science in Mathematics

Overall Average GPA: 3.65/4.3, Top 19%
Minor in Economics

• Specializations:
Business Analytics (15 credits, GPA: 4.18/4.3)
• Relevant Course:
Analysis, Probability, Algebra, Differential Equation, Geometry, Statistical Inference, Numerical Analysis, Mathematical Finance, Cryptography, Computer Programming, Principles of Economics, Microeconomics, Macroeconomics, Statistics, Econometrics

University of California, Berkeley

Summer Sessions
Overall Average GPA: 3.65/4.0

PROFESSIONAL EXPERIENCES

Data Analyst Intern

Taelor | San Francisco, US

• Conducted cohort analysis using SQL and Python to support business decision-making processes, resulting in a 5% increase in operational efficiency; Analyzed over 3,000 preview messages to identify key engagement metrics, leading to a 12% improvement in targeted communication strategies.
• Created visualization dashboard of persona using Looker and Google Analytics to represent business metrics for quarterly reviews, which were commended for their clarity and insights.

Data Scientist

Zomma Protocol | Taipei, Taiwan

• Conducted 10+ DeFi derivative secondary analysis reports by tracking macroeconomic data, analyzing industry cycles, and extracting evaluation indicators, etc. to provide feasibility analysis and persuade CEO to initiate the project and sustain improvements, significantly impacting business strategies.
• Automated extraction and pre-processing of over 50k financial data on a weekly basis by writing Python script to use Deribit API; Saved 60% work time by engineering ETL pipeline to automatically store data into MySQL database for further analysis.
• Developed “Two-Stage Estimation” for DeFi option pricing using Black-Scholes, Heston model calibration, and Monte Carlo simulation in Python, improving real-time pricing accuracy; Increased trading volume by 20% and market liquidity by 15%, reported to CTO influencing trading algorithm enhancements, and prepared for academic journal publication.
• Created an automatic and stable dynamic hedging mechanism using Auto Market Maker principles, considering impermanent loss and liquidity, while ensuring Delta Neutral and effective control over Vega Risk in Python; Reduced hedging costs by 12%, and increased trading liquidity of the platform by 10%; Reported the results to the CEO, leading to its implementation in the beta version, continuing to operate successfully to date.

Research Assistant

Dept. of Quant. Finance, National Tsing Hua University | Hsinchu, Taiwan

• Reproduced 10+ quantitative financial models in Python by reading pioneer research papers and reports in top journals, understanding underlying mathematic implementation. Demonstrated strong understanding of current FinTech trends and the latest developments in the academic area.
• Designed a premium formula for DeFi derivatives to minimize the buy-sell spread and facilitate transactions, increasing market liquidity theoretically and outperforming industry standards at the time; Discussed on-chain fund possibilities with professors and top security company analysts, who responded positively to its innovative approach.
• Organized academic conferences weekly for 500+ researchers and investors, demonstrating cutting-edge FinTech innovations; Showcased leadership by building connections and fostering discussions, gaining insights into the latest innovations; Increased visibility for the team and enhanced external recognition. High-rated, educational presentations significantly benefited the audience.

Data Scientist

Fubon Financial Holding Co., Ltd. | Taipei, Taiwan

• Employed over-sampling and SMOTE to address data imbalance, improving data quality for later usage. Conducted feature engineering and Exploratory Data Analysis (EDA) on credit card transaction data, generating new features that boosted model performance by 25%.
• Developed tree-based models (Random Forest and XGBoost) using Python to predict repurchase intensity scoring by over 2000 features of financial customers data, enabling more targeted marketing strategies; Reported results to the manager, leading to the implementation of personalized customer retention campaigns, boosting repurchase rates by 13% in the following months.
• Implemented diverse variable selection methods (PCA and WOE), and improved model performance by implementing various model optimization methods, including data resampling, feature engineering, K-fold cross validation, etc., enhancing predictive accuracy from 60% to 83.89% and achieving 0.68 AUC.
• Interpreted feature importance using SHAP values to identify key features, providing clear visualization analysis to demonstrate each feature's contribution; Presented findings to both technical and non-technical partners, facilitating quick understanding. The positive feedback led to plans for integrating these insights into business strategies, aiming to enhance data-driven decision-making processes.

Undergraduate Teaching Assistant

National Taiwan University | Taipei, Taiwan

• Performed all assistant teaching duties, including mentoring, and lecturing.
• Graded 30+ students' exams. Skilled in research and materials development in mathematics.

PROJECTS

Diffusion Model for MNIST Data

Python/Diffusion Model/Deep Learning/PyTorch/Hyperparameters Tuning

• Implemented Denoising Diffusion Probabilistic Models paper, employing UNet architecture to predict conditional mean and error term, achieving clearer image synthesis with the error term model.
• Demonstrated model differences in output, sampling process, and loss function; Model size: 4.3 MB; Training time: 20 secs per epoch.

Optimizing Sentiment Analysis: An Alternative Way to Revisit SESTM Algorithm

Python/MatLab/Big Data/Sentiment Analysis/Text Mining/Hyperparameters Tuning

• Revisited the text-mining methodology, Sentiment Extraction via Screening and Topic Modeling (SESTM), that extracts information from 25M unique news articles from Dow Jones News to predict asset returns.
• Conducted the hyperparameter tuning to achieve higher accuracy, and tried to find the Benchmark performance of the SESTM algorithm. Fine-tuned SESTM-based sentiment analysis model earns an annualized Sharpe ratio of 5.02 over 4.21, marking a significant advancement in professor-led research.

Big Data - Loan Default Analysis

R/Big Data/Regression Modeling/PCA/K-means/LASSO/Machine Learning/Neural Network

• Conducted in-depth EDA to credit card customer data (30K records) with Python and MySQL for data-driven decision-making
• Built ML models (logistic regression and random forest) to predict the loan default rate, utilizing features selection techniques (LASSO and PCA), with accuracy 82.09%, uncovering the key factors influencing the likelihood of loan default
• Implemented clustering techniques (Kmeans and Hierarchical Clustering) to effectively identify significant predictors feature importance, enhancing segmentation strategies

Automatic Market Making(AMM) Research Project

AMM/Liquidity Mining/DeFi/Constant Product Markets/Geometric Mean Market (G3Ms)

• Designed a set of automatic market making mechanisms and the realization of liquidity mining with smart contracts. Empirically analyzed the data on the blockchain to present the simulation of the market-making process.
• Applied some dynamic equilibrium model methods to the problem of decentralized finance.

Big Data Marketing Analysis

SPSS/Big Data/RFM Analysis/CAI/CRI/Recommendation System/Customers Clustering

• Used a data set that recorded 4481 supermarket transaction records from Jan 2009 to Dec 2010, to conduct big data marketing analysis, including RFM, CAI, CRI indicator analysis, basket and clustering analysis, utilizing SPSS.
• Gave every type of customers particular scores of purchasing behaviors, letting company utilize my result.

Text Mining DNN Model Analysis

Python/Text Mining/Machine Learning/DNN/Pytorch/Sklearn/Time Series

• Utilized text mining techniques (Tf-Idf, word2vec), and BERT pretrained model to construct reasonable word embedding
• Built a deep neural network (DNN) model with PyTorch to predict Australia's 30 Year Government Bond yield using a dataset of 1M Australian news headlines spanning 17 years, achieved 95% accuracy with 0.005 MSE after hyperparameters tuning

Telecom DCB Service Optimization

R/Regression Modeling/RFM Analysis/Generalized Linear Models/Customers Clustering

• Constructed a robust Generalized Linear Model (GLM) using R to optimize the performance of the DCB service for T STAR, a prominent Taiwanese telecommunications company.
• Effectively contributed to the organization's growth by formulating and presenting a comprehensive marketing strategy, leveraging valuable insights obtained through the rigorous analysis facilitated by the GLM.

Specific Application of Copula Model, Loss Distribution Approach model (LDA)

• Presented how dependence between risks is important for a fair quantification of the economic capital and introduced the use of copulas to model this dependence.

TECHNICAL STRENGTHS

• Programming Languages: Python, R, SQL
• Machine Learning Framework: PyTorch, TensorFlow, Scikit-Learn
• Softwares: Stata, MATLAB, SPSS
• Machine Learning Algorithms: Classification, Clustering, Neural Networks, Natural Language Processing, Large Language Model
• Statistics: Time Series, Regression, A/B Testing, Bayesian Modeling, Causal Inference
• Data Visualization: Looker, Tableau, PowerBI, MS Excel
• Other Tools: Git, Jupyter, GCP (Big Query), PySpark, Blockchain

EXTRA CIRRUCULARS

• Finished mandatory military service in Taiwan in 2022 spring.
• Attended Taiwan Fintech Forum in 2021. Presented Onchain Fund application in campus fintech creative.
• Attended Meichu Hackathon in 2019. Constructed an auto indoor parking positioning system.
• Participated in NTU Azalea Festival 2018. Contributed to department expo.
• Participated in Math Camp 2018 organized at National Taiwan University.

AWARDS & CERTIFICATIONS

• Career Essentials in Generative AI by Microsoft and LinkedIn - Microsoft
• INSIDE LVMH Certificates - LVMH
• Google Analytics Individual Qualification - Google
• 1st in Mathematics in the Regional Science Fair and the "National Best Local Teaching Material Award" in the National Science Fair

LANGUAGE SKILLS

• Mandarin Chinese (native)
• English (fluent)
• Taiwanese (dialect)