Pfizer Externship via ExternNortheastern M.S. student applying statistics, machine learning, and mathematical modeling to real-world problems.
Selected projects in predictive modeling, statistical analysis, and numerical computing from my graduate and undergraduate studies.
Northeastern University · May 2026
Built and compared regression models to predict listing prices using Boston Airbnb data. The project combined data cleaning, feature engineering, model selection, and evaluation in Python.
Northeastern University · May 2026
Benchmarked least-squares methods to understand how numerical stability and computational cost change under different data conditions.
Northeastern M.S. student applying statistics, machine learning, and mathematical modeling to real-world problems.
I am pursuing an M.S. in Applied Mathematics with a Data Science concentration at Northeastern University, after earning B.S. degrees in Mathematics and Statistics at UC Davis. I have analyzed business data during an internship, taught mathematics, and built academic projects using Python, R, SQL, and MATLAB. I enjoy translating quantitative results into clear, useful insights.
Externships
Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship
In progress
Experience
Math Instructor
Think Academy · Jan 2025 - Sep 2025
Network Engineering Intern — Data Analytics & AI
Beijing Yintong Chuangda Technology Co., Ltd. · Jun 2023 - Sep 2023
Education
Northeastern University
M.S. in Applied Mathematics, Data Science concentration · Class of 2027
University of California, Davis
B.S. in Mathematics and B.S. in Statistics · Class of 2024
Skills
Northeastern University · May 2026
Built and compared regression models to predict listing prices using Boston Airbnb data. The project combined data cleaning, feature engineering, model selection, and evaluation in Python.
Built and compared regression models to predict listing prices using Boston Airbnb data. The project combined data cleaning, feature engineering, model selection, and evaluation in Python.
Processed 4,419 raw listings into 3,466 analytical records and constructed 63 encoded predictors. Compared Random Forest, Gradient Boosting, XGBoost, and ridge regression; tuned Gradient Boosting with five-fold cross-validation.
Evaluated models on a 694-record test split using RMSE, MAE, and R², and documented feature importance, residual diagnostics, and preprocessing choices in a notebook and report.
Northeastern University · May 2026
Benchmarked least-squares methods to understand how numerical stability and computational cost change under different data conditions.
Benchmarked least-squares methods to understand how numerical stability and computational cost change under different data conditions.
Compared normal equations, explicit inversion, QR, SVD, and NumPy least squares on simulated 2,000 × 500 matrices. Tested conditioning, noise, numerical precision, and ridge regularization.
Visualized solution error and computational cost to explain the tradeoffs among methods.
University of California, Davis · Jun 2024
A team project exploring vehicle insurance claim classification using a dataset of 58,592 records.
A team project exploring vehicle insurance claim classification using a dataset of 58,592 records.
Contributed data preprocessing and exploratory modeling. Examined numerical, categorical, and binary predictors, feature importance, correlations, and dimensionality reduction; the team tested random forest classification and imbalance-handling methods.
Documented experiments and analytical outputs in notebooks and a project report, connecting vehicle and customer attributes to the claim-prediction task.
Northeastern University · Dec 2025
Prepared and explored a modified obesity teaching dataset to practice reproducible data cleaning and statistical analysis in R.
Prepared and explored a modified obesity teaching dataset to practice reproducible data cleaning and statistical analysis in R.
Imputed missing height and age using group-specific summaries, filtered weight outliers, derived BMI, and retained 2,094 records for analysis.
Produced category summaries, confidence intervals, and overall and stratified correlations; recorded transformations and results in an R Markdown notebook.
Northeastern University · May 2026
Evaluated how interpolation methods affect the quality of reconstructed photographs at different downsampling factors.
Evaluated how interpolation methods affect the quality of reconstructed photographs at different downsampling factors.
Compared nearest-neighbor, bilinear, bicubic, and Lanczos interpolation with MATLAB and Python. Measured RMS error, PSNR, and edge-region error, and generated figures and CSV outputs.
On one tree-image crop at 0.25 downsampling, bicubic interpolation achieved 29.07 dB PSNR versus 26.04 dB for nearest-neighbor. Presented the relationship between interpolation choice, detail, and error.
Northeastern University · Dec 2025
Contributed to a team practicum translating hospital facility, bed-category, and capacity data into a relational database.
Contributed to a team practicum translating hospital facility, bed-category, and capacity data into a relational database.
Developed an R and SQL schema with three related tables, primary and foreign keys, and fields for licensed, staffed, and census bed counts.
Added schema-inspection and row-count checks to support review of the database structure and loaded data.
✅ Verified by Extern · ⏱️ In progress
Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.
The project prototyped an AI-powered document intelligence pipeline that combined optical character recognition, language models, and retrieval-augmented generation to process enterprise PDFs. The work researched LLM architectures and capabilities, designed extraction workflows, and produced demonstrative prototypes showing automated data extraction from sampled documents.
.jpg)
I researched core LLM concepts and explained how they function, documenting model architectures, training dynamics, and capabilities relevant to document understanding and extraction.
