Aetna at CVS Health:

Data Scientist  Jan 2022 - Jul 2022

• Design and implement the Monte Carlo simulation pipeline to obtain range estimates of member disenrollment upon product updates. Enhance efficiency through parallelization and vectorization. Complete 100K+ simulations on datasets with millions of rows within an hour on a distributed system.
• Investigate optimal campaign test design and calculate the efficient allocation of sample sizes (with given power and significance level) using quadratic programming.
• Implement a nested parallelized multivariate risky cohort identifier completely from scratch using the Dask backend, catering to imbalanced datasets.
• Investigate drivers of member lapse using ad-hoc analysis and develop a member-churn data warehouse on Hive.

Data Scientist (Part-Time)  Oct 2021 - Dec 2021

• Build supervised machine learning models to forecast sales call receptivities of Medicare enrollees.
• Automate and engineerize the data cleaning, modeling, hyper-parameter tuning, visualizations, and the report summary pipeline.
• Develop the ETL pipeline on Hive database to get prepared for modeling.

Data Scientist (Internship)  Jun 2021 - Aug 2021

• Automate data collection pipeline from web portals using Javascript, saving 100+ analyst hours. Expedite the whole process by multi-threading using asynchronized code.
• Conduct n-gram analysis (TF-IDF), topic modeling (LDA), and sentiment analysis on thousands of sales call transcripts.
• Predict clients' dis-enrollment status by leveraging supervised machine learning models (logistic regression, k nearest neighbors, random forest, gradient boosting, support vector machine, adaboost, neural networks, etc.).
• Experiment audio data preprocessing scopes (ambient noise, volume, etc.) for call recording transcription.
• Generate business insights and actionable recommendations from modeling and analysis results.

Percolata:

Machine Learning Engineer (Part-Time Internship)  Sep 2020 - Jun 2021

• Collaborate with machine learning engineers to build algorithmic trading applications on Google Cloud Platform in an AGILE environment. Deploy machine learning models to google cloud applications and ai-platform.
• Design and implement reinforcement learning stock triggers using Pytorch. Explore machine learning models (Linear Regression, Auto Arima, Random Forest, XGBoost, CatBoost, etc) for stock price forecasting. Assess the significance of numerous stock indicators (sentiment scores from the latest news, RSI, Stoch, MACD, VWAP, etc) for stock price forecasting.
• Design a heuristic stock screener based on Ornstein-Uhlenbeck based mean reversion property, trend, and volatilities of potential hedging portfolios, slashing over 95% of execution costs from the brute-force stock screeners.
• Code up web-crawlers using BeautifulSoup and Selenium to obtain financial news and stock tickers recommended by the Motleyfool.
• Design and implement cloud-based CI/CD workflow to automate the testing and deployment of source code.
• Implement google cloud functions and cloud run services to aid the stock trading evaluations and the interactions with Alpaca API, Interactive Brokers API, and google cloud API.

Institute for Health Metrics and Evaluation:

Research Assistant  Feb 2019 - Sep 2019

• Researched and executed Iterative Filtering, Least Square Method, and Particle MCMC in R to position parameters in differential equation systems. Conducted tests with 10+ algorithms to forecast traveling flows across territories within Equatorial Guinea. Accelerated data cleaning and model training process by vectorizations using linear algebra.
• Saved 80% storage space for repeated simulations by wrangling data output format. Implemented ETL to prepare data across different sources for model training. Employed statistical tests to expedite posterior inference to assess the model performance. Utilized ggplot to create plots to foresee model functionality. Used Raster and MapTools to plot commuting flows on a map.