Currently an Extern @Pfizer·🟢 Open to work
I'm a junior at UTA, focused on Software Engineering and excited to create innovative tech solutions.
Explore my portfolio featuring innovative projects, including AI-driven solutions and document automation with Pfizer.
.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.
Implemented and tested NestJS REST API functionality with MongoDB/Mongoose, DTO validation, consistent error handling, and reusable controller-service architecture; wrote Jest unit tests for backend services including mocked Mongoose operations; used Postman and local admin/public clients to verify API contracts; collaborated through GitHub pull requests, code reviews, rebasing, CI checks, and merge-conflict resolution.
I'm a junior at UTA, focused on Software Engineering and excited to create innovative tech solutions.
I'm a junior Software Engineering student at the University of Texas Arlington, eager to kickstart my career. I'm passionate about leveraging technology to solve real-world problems, particularly through AI and automation.
Externships
Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship
Pfizer
Experience
Backend Developer
VetRanker — Veterinary Ranking Platform · 2025 - Present
Software Engineering Coursework — University of Texas at Arlington
University of Texas at Arlington · 2024 - Present
Personal Project
Full-Stack Internship Tracker · 2026
Full-Stack Developer
Doctor Appointment Booking Web Application · 2025
Education
The University of Texas at Arlington
B.S. in Software Engineering · Class of 2028
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 prototyped AI approaches for reading pharmaceutical documents using OCR, computer vision, and large language models. Work documented how tokenization, transformer self-attention, and next-token prediction enable text understanding, and it explained pretraining plus fine-tuning with human feedback and reinforcement learning.
I produced a technical explanation of how LLMs process pharmaceutical text, detailing tokenization, transformer self-attention, next-token prediction, and the roles of pretraining, fine-tuning with human feedback, and reinforcement learning.
I explained how LLMs tokenized text, used transformer self-attention to model long-range relationships, and generated next-token predictions. I described pretraining on large corpora and subsequent fine-tuning with human feedback and reinforcement learning.
The submission showed data cleaning with Pandas, JSON normalization into a flat DataFrame, text standardization scripts, and image preprocessing using OpenCV/PIL. It documented steps, code, and a cleaned image result produced in Google Colab.
During the externship I extracted text and bounding boxes from a three-page SDF, used regex and PyMuPDF search to locate labels, and parsed manufacturing and expiration dates. The submission produced date lists and noted table layout issues.
I compared three OCR engines on a scanned pharmaceutical certificate. I ran each tool, inspected outputs for lot number, dates, product codes, tables, and special characters, and found Tesseract produced the cleanest, most organized text; PaddleOCR and EasyOCR had more character and layout errors.
I built a RAG pipeline that ingested a PDF, used sentence-transformer embeddings and sentence-based chunking (512 tokens, 50 overlap), combined vector search with BM25, applied a SentenceTransformer reranker, and produced targeted answers to document queries.
I compared three RAG configurations (varying top_k, similarity threshold, and a reranker) against an SDF query. I logged retrieved chunks, short answers, and confidence scores, and found the baseline (top_k=5, no threshold, no reranker) returned the most accurate QC test methods.
The project processed multi-page PDF bundles, extracted page text and metadata, used page-to-page context and prompts to detect document boundaries, and produced a JSON mapping pages to doc_id and doc_type (e.g., Certificate of Quality, Packaging Specification).
The project processed uploaded PDFs, extracted text (with OCR fallback), chunked content with metadata, stored embeddings in FAISS, and routed queries by document type before sending context to an open-source LLM. The deliverable included a working Gradio UI, retrieval logic, and a slide/doc that…
Implemented and tested NestJS REST API functionality with MongoDB/Mongoose, DTO validation, consistent error handling, and reusable controller-service architecture; wrote Jest unit tests for backend services including mocked Mongoose operations; used Postman and local admin/public clients to verify API contracts; collaborated through GitHub pull requests, code reviews, rebasing, CI checks, and merge-conflict resolution.
Built a React-based application with a Node.js/NestJS and MongoDB backend for creating, updating, deleting, and tracking internship applications and statuses; debugged frontend-backend integration issues and aligned schemas and UI behavior across the stack.
Developed a responsive React frontend and NestJS/MongoDB backend for appointment scheduling and patient record management; implemented authentication using JWT tokens and bcrypt password hashing and organized backend logic using NestJS modules, controllers, and services.
