Pfizer
AI document insights & data extraction
A learning-focused externship exploring how to make complex documents easier to search and verify. I developed a document-processing and retrieval prototype and examined where extraction, retrieval and answer generation could fail.
Timeline
July–October 2026 · Pfizer Advanced: AI-Powered Document Insights & Data Extraction, delivered through Extern.
My contribution
Developed OCR and document-processing workflows, organized document classification and routing, and built a Gradio interface for retrieval-based questions. Used source-page references to make answers easier to review.
How I worked
Separated document extraction, retrieval and answer generation so each stage could be assessed. Explored chunking, semantic search and prompt design, then documented limitations and possible improvements.
Work & learning
A working prototype and practical experience evaluating document AI. The public notebook shows the implementation approach; it is a learning prototype rather than a claim of production deployment.
Related work


Document search with verifiable sources
A Python document-search prototype combining OCR, semantic retrieval and a Gradio interface with source-page references.