I summarized how large language models (LLMs) work, explaining their transformer architecture, self-attention, training by next-token prediction, and fine-tuning with human feedback.

I prepared and cleaned pharmaceutical documents using Python in Google Colab, applying data- and image-processing techniques to improve OCR readiness.

I converted multi-page SDF PDFs into structured, field-level outputs using PyMuPDF and regex parsing.

I compared Tesseract, PaddleOCR, and EasyOCR on a pharmaceutical supplier PDF and reported which engine performed best for layout-aware extraction.

I built and tested a complete RAG pipeline for document retrieval using LlamaIndex, including embeddings, chunking, hybrid retrieval, and reranking.

I compared three RAG retrieval configurations on a sample SDF document and logged how retrieval settings affected answers.

I built a page-level segmentation and classification output for a bundled PDF, producing structured metadata and document-boundary labels.

I built a Gradio chatbot interface and connected it to a document-backed RAG pipeline, adding UI features and live PDF Q&A.