Wayfair Externship via ExternI'm passionate about harnessing AI to transform data extraction and insights. Let's innovate together!
Dive into my portfolio showcasing AI-driven projects in data extraction and insights that aim to make a difference.
.jpg)
Pfizer · Sep 2026 · ✅ Verified by Extern
Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.
Designed and implemented a transactional e-commerce backend over a 1,000+ product PostgreSQL database, supporting 50+ concurrent checkout simulations without failure. Built 10+ RESTful API endpoints for cart, checkout, and order management, achieving full multi-step purchase workflows with < 200ms average query latency through indexed queries and optimized joins. Implemented atomic transaction logic for inventory updates and balance transfers and reduced relational schema from 6 to 4 tables while preserving historical prices.
Investigated trade-offs between transfer learning and task-specific modeling by fine-tuning a pre-trained TGCN and comparing with a custom 1D-CNN. Achieved 55.05% accuracy (F1: 0.87) on a 5-class dataset by training a 1D-CNN on 100+ labeled videos using MediaPipe keypoints. Reduced inference latency to 20.33ms/sample by fine-tuning a TGCN and benchmarked models performing error analysis.
I'm passionate about harnessing AI to transform data extraction and insights. Let's innovate together!
I am currently engaged in an externship with Pfizer, focusing on AI-powered document insights and data extraction. This experience allows me to refine my skills in leveraging technology to enhance data handling and analysis.
Externships
Wayfair n8n AI Agent Engineering Externship
Wayfair
Experience
Internship - Frontend Developer & Product Designer
Liberata · Aug 2025 - Present
Education
Duke University (Durham, North Carolina, USA)
Bachelors of Science. Computation and Design, Computer Science (DKU) · Class of 2025
Duke Kunshan University (Kunshan, Jiangsu, China)
Bachelors of Science. Computation and Design, Computer Science (DKU) · Class of 2027
Skills
✅ Verified by Extern
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 work examined how AI ingests and extracts data from pharmaceutical documents, combining OCR, computer vision, and LLM mechanics. Deliverables included a technical summary of LLM internals, runnable Python notebooks for data cleaning and image preprocessing, PyMuPDF-based extraction pipelines that produced structured field records, and an OCR comparison that identified the most accurate engine.
I produced Python notebooks and extraction examples that documented LLM internals, implemented image preprocessing for OCR, extracted word-level text and bounding boxes with PyMuPDF, applied heuristics to produce structured field records, and compared Tesseract, PaddleOCR, and EasyOCR.
Designed and implemented a transactional e-commerce backend over a 1,000+ product PostgreSQL database, supporting 50+ concurrent checkout simulations without failure. Built 10+ RESTful API endpoints for cart, checkout, and order management, achieving full multi-step purchase workflows with < 200ms average query latency through indexed queries and optimized joins. Implemented atomic transaction logic for inventory updates and balance transfers and reduced relational schema from 6 to 4 tables while preserving historical prices.
Investigated trade-offs between transfer learning and task-specific modeling by fine-tuning a pre-trained TGCN and comparing with a custom 1D-CNN. Achieved 55.05% accuracy (F1: 0.87) on a 5-class dataset by training a 1D-CNN on 100+ labeled videos using MediaPipe keypoints. Reduced inference latency to 20.33ms/sample by fine-tuning a TGCN and benchmarked models performing error analysis.
