AI-Based IP Law Textbook Generation (Sep 2024-May 2025)

Led the end-to-end automation pipeline for converting and extracting content from 100+ unstructured PDFs and PPTs into structured Word documents using PowerShell, PyMuPDF, and Python-docx.
• Extracted and tagged text, headings, and images; performed word count validation to ensure fidelity to original legal content.
• Used OpenAI API to summarize extracted material and generate keywords; compared content against the client’s Target TOC using Hugging Face’s Sentence Transformers for semantic alignment.
• Created section-wise Word documents for each TOC topic to centralize relevant content, improving navigability and reusability for downstream tasks.
• Designed and deployed a robust Retrieval-Augmented Generation (RAG) workflow using Llama 3, and OpenAI API, enabling intelligent content rewriting and enrichment.
• Built and deployed a live Streamlit app to empower agents with on-demand retrieval and rewrite capabilities, improving rewriting speed and contextual accuracy by over 70%.
https://www.twine.net/signin