I explained how large language models learn and how transformer attention enables contextual prediction.

I prepared and cleaned pharmaceutical-style documents using Python in Google Colab, applying Pandas and image preprocessing to improve OCR readiness.

I extracted and structured text from multi-page SDF PDFs using PyMuPDF, then identified and normalized key fields like dates and vendor cues.

I compared Tesseract, PaddleOCR, and EasyOCR on a scanned pharmaceutical certificate and reported comparative metrics and recommendations.

I built and tested a complete RAG pipeline using LlamaIndex, implementing hybrid retrieval, reranking, and page-level chunking for document search.

I ran comparative RAG experiments to test three retrieval configurations and logged how chunking, top-k, thresholds, and reranking changed returned context for SDF queries.

I built a page-level segmentation and metadata extractor that identified document boundaries and labeled each page by document type.

I built a themed Gradio chatbot interface with multiple reading modes and UX features for document Q&A.

I compiled a running 'Holiday Game Plan' document to track Week 1 analysis, inventory, and prompts for querying Breaking Games' six data sources.

I designed dim_product, connected tables with keys, joined customer-event tables, and produced a product performance summary to inform a reorder recommendation.

I built a Q4 ad allocation plan linking $37K in Meta spend to revenue and campaign-level ROI, and recommended which campaigns to keep, cut, or grow.

I generated and tested hypotheses about why carts were abandoned, identifying shipping cost and potential bot patterns in the checkout export.