I explained how large language models work, describing their training, token-by-token generation, and common failure modes.

I prepared and processed pharmaceutical-style documents with Python in Google Colab, covering data cleaning, JSON normalization, text standardization, and image preprocessing for OCR.

I extracted text and structured fields from multi-page SDF PDFs using Python and PyMuPDF, producing an executable Colab notebook with the extraction workflow.

I compared Tesseract, PaddleOCR, and EasyOCR on scanned pharmaceutical supplier documents and produced a quantitative comparison and recommendation.

I built and evaluated a complete RAG pipeline for document retrieval, documenting design choices and performance findings.

I compared three RAG retrieval configurations on a sample SDF document and logged differences in retrieved chunks, best excerpts, and answer quality for a QC-related query.

I built a page-level segmentation and metadata extraction pipeline for bundled regulatory PDFs, producing per-page doc_type and page_in_doc labels.

I built a Gradio-based document Q&A chatbot interface and integrated retrieval features to return context-aware answers from uploaded PDFs.

I built a Prompt Node for the Moodboard Generator, documenting example prompts and node configuration.

I built an end-to-end trend discovery deliverable for the Area Rug category, packaged as a downloadable report and supporting workflow screenshots.

I built a Competitor Monitoring AI Agent that fetched, merged, and analyzed rival product listings for area rugs, then generated a PDF benchmarking report.

I updated the AI Insights & Content Agent to expand its outputs and inputs into a multi-channel content plan aligned with brand voice.