I explained how large language models (LLMs) work, summarizing their architecture and training process.

I prepared and cleaned pharmaceutical-style documents with Python, using Colab notebooks to process data, clean text, and preprocess images for OCR.

I extracted word-level text and bounding boxes from a multi-page SDF using PyMuPDF and produced a structured JSON output.

I compared Tesseract, PaddleOCR, and EasyOCR on a scanned pharmaceutical supplier PDF and documented which engine handled layout and field extraction best.

I built and evaluated a RAG pipeline using LlamaIndex to parse a technical PDF, create embeddings, run hybrid retrieval with reranking, and generate answers to certificate-focused prompts.

I compared three RAG retrieval configurations on a pharmaceutical SDF PDF and reported which setup retrieved the most relevant chunks and produced the best answers.

I built a page-level classifier that split a 10‑page pharmaceutical PDF into 8 logical documents, labeled each page with doc_id and doc_type, and exported the results as assignment-ready JSON.

I built a Gradio document Q&A chatbot that processed PDFs, indexed chunks, and returned retrieval-grounded answers with UI controls and citation badges.