I summarized how large language models function and the role of tokenization and transformer architecture in generating text.

I prepared and cleaned a raw pharmaceutical dataset using Pandas in a Google Colab notebook.

I extracted and structured text from multi-page SDFs using Python, then produced a shared document with the parsed output.

I compared Tesseract, PaddleOCR, and EasyOCR on a scanned pharmaceutical supplier PDF and documented differences in layout handling and text extraction.

I built and tuned a Retrieval-Augmented Generation pipeline using LlamaIndex, producing a runnable notebook and supporting document.

I compared Gemini, Mistral, and Phi-2 in a RAG pipeline and measured retrieval and generation differences across configurations.

I built a segmentation and metadata pipeline to split bundled PDFs into page-level chunks and tag each chunk for routing and extraction.

I built a Gradio-based chatbot UI and connected it to a local RAG pipeline to answer questions from a PDF, packaging the interface and prompt handler as separate functions.