Pfizer Externship via ExternSophomore data science major exploring applied analytics, modeling, and data-driven projects as I prepare to launch my career.
Portfolio of class projects, personal analyses, and early practical work that show my approach to data cleaning, modeling, and visualization.
.jpg)
Pfizer · ✅ 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.
Sophomore data science major exploring applied analytics, modeling, and data-driven projects as I prepare to launch my career.
I am Borun Li, a sophomore studying Data Science at Duke Kunshan University and Duke University. I work as a Data Analyst and assist research in neuroscience, where I apply Python, R, MATLAB, and statistical modeling to data cleaning, exploratory analysis, figure design, and rapid reporting. I have experience as a research assistant and a full-time chemistry tutor, and I focus on building reproducible pipelines, feature engineering, and clear technical writing while I continue learning and gaining hands-on research experience.
Externships
Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship
In progress
Experience
Student Research Assistant to Dr. Chia-Chien Chen, Assistant Professor of Neuroscience
Circuit and Behavior Lab (Chen Lab) at DKU · May 2026 - Present
Full-time Chemistry Tutor
Kunshan Gaozong CEE Education and Training Center Co., Ltd · June-Sep 2026
Intern Research Assistant, Innovation Department
Neuroscience Lab at Care YaYa · July-Aug 2026
Education
Duke Kunshan University (DKU) & Duke University
B.S. in Data Science (Intended) · Class of 2029
Skills
✅ 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 externship explored AI-driven document intelligence for enterprise PDF workflows using OCR, large language models, and retrieval-augmented generation. The work reviewed LLM fundamentals and documented how those components fit together for an end-to-end prototype. Deliverables included written explanations and technical notes about model behavior and integration points.
I inspected a 39-page scanned PDF spanning nine document types, catalogued OCR errors and mixed date formats, and documented specific extraction risks such as wrong date parsing, repeated headers, redactions, and vendor-name confusion.
I inspected a 39-page scanned PDF containing roughly nine document types, identified mixed date formats and ambiguous ages, noted OCR errors on table text, and listed risks: wrong date parsing, repeated headers, mixed document types, redactions, and multiple vendor names.
I received messy document files, identified missing values and encoding issues, then wrote Python notebooks in Colab to parse nested JSON, clean text (ftfy, regex, Unicode normalization), fix missing values with pandas, and apply image denoising and thresholding for OCR-ready outputs.
The task presented messy SDF PDFs. I ran PyMuPDF to extract words and bounding boxes, inspected outputs for tables and special characters, and recorded where text garbled or columns merged. I produced a note of extraction issues and suggested regex/anchor strategies for dates.
