Proposed and executed two forecasting frameworks to predict the performance of 1330 mutual funds using models such as long and short-term memory networks and feedforward neural networks with time-series data and alternative data.

Collaborated with team members to obtain a clear conclusion of the linearity between prediction performance and actual results, returning a remarkably small out-of-sample root mean squared error of 0.05.

Assessed an implicit attention mechanism based on random feature maps working with a sixth-power kernel; Led a team of six to deliver concentration results for it.

Proposed two types of random feature maps for kernel linearization, both unbiased with a reasonable error term.

Managed the Performer method to accurately and practical estimate regular full-rank attention.

Prepared and managed data extraction, transformation, and documentation for a large waste management study which consists of thousands of pdf records ( Landfill Annual/Quarterly Reports).

Extracted data from pdf files into CSV files in Python, conducted data cleaning, imputed missing data.

Understand the CDW material flows and complete the data visualization of the patterns of recycled materials that are available for reuse on construction projects in Tableau.

Led the data analysis team to execute research on the impact of Airbnb short-term rentals, it involved close cooperation among members to share information and develop solutions to meet broad array of deliverables.

Performed data preprocessing, cleared out-of-range values and missing values, increased the quality of data. Designed a new database to manage daily Airbnb transaction data, organized and reviewed more than a million data points to locate required data for entry, and converted data workflows from Python.

Processed data integration and constructed a model to assess common features based on historical Airbnb price data using the above-mentioned scripts. Analyzed and developed technical and functional specifications for databases.

Participated in predictive modeling, tested multiple classification models such as Random Forest and regression models, and performed hyperparameter tuning to optimize model predicting power in Python.