Pfizer Externship via ExternMaster's student building practical AI and document data extraction skills through hands-on externship work to boost my resume and readiness for internships.
A portfolio of course work and the Pfizer Advanced externship that documents my hands-on work with AI models, document insights, and data extraction techniques.
.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.
Master's student building practical AI and document data extraction skills through hands-on externship work to boost my resume and readiness for internships.
I am Kevin Pasato, a graduate student in Computer Engineering at Stevens Institute of Technology pursuing a masters degree. I have limited professional experience and I am focused on building practical skills and strengthening my resume. I completed the Pfizer Advanced: AI-Powered Document Insights & Data Extraction externship to gain hands-on experience with AI and data extraction workflows.
Externships
Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship
In progress
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 project surveyed large language model architectures and applied them to document intelligence tasks, mapping how OCR, RAG, and model selection interact in extraction pipelines. The work compared model capabilities and integration patterns, producing a synthesized explanation of LLM mechanics as they apply to enterprise PDF workflows.
I produced a concise summary of LLM training dynamics, including next-word prediction, weight updates from errors, and the role of human ratings, and described why LLMs can generate confidently incorrect outputs.
The work summarized how LLMs train by predicting the next word from massive text, how errors adjust model weights, how human ratings refine responses, and why models can produce confident but incorrect answers.
