Welcome to Document AI: From OCR to Agentic Doc Extraction, built in partnership with LandingAI where I'm executive chairman. A lot of data is locked up in PDF files and other documents on our laptops, on the web, and in company's cloud storage. In this course, you'll implement document processing pipelines that transform your complex documents into LLM-ready markdown text and extracts information. for analysis. You begin by explaining traditional OCR which only extracts text and then add techniques to detect document structure and identify visual components such as tables with complex formatting like merged cells or charts with captions or forms with checkboxes. You'll implement an agentic workflow that combines layout detection with LLM-based reasoning. Then, you'll learn to use Agentic Document Extraction, or ADE, a tool designed by LandingAI that automates this entire workflow for you. I'm delighted that the instructors for this course are David Park, who's Senior Director of Applied AI, and Andrea Kropp, who is an Applied AI engineer. Both have helped many developers build complex document AI systems. Thanks Andrew. We're excited to teach this course together. Traditional OCR systems worked by segmenting documents into individual characters and then classifying each character using supervised learning. These systems could extract characters accurately, but they lacked understanding of how different parts of a document hang together, such as the document structure or tables or forms, or even what is the reading order of different elements of a document. Landing AI's ADE takes a different approach. It treats each page of a document as an image and breaks that image down into parts and then extracts from each part separately using an agentic workflow. ADE treats documents as visual objects in which meaning is encoded in layout structure and spatial relationships. It uses custom visual models that directly interpret complex tables, graphs, images, charts, and other elements. and grounds every extracted piece of text to a precise location on the page. On top of this, ADE uses an agentic orchestration layer that decomposes complex documents into smaller sections for careful examination in several steps. Humans don't process a document with a single glance. Instead, we examine the different parts of the document to pull out information piece by piece in multiple iterations. ADE works the same way. For example, given a complex document, ADE it might extract a table and then further extract the table structure, identifying rows, columns, merged cells, and so on. In this course, you'll explore traditional OCR-based processing pipelines to see what they can do and also their limitations. You'll also build your own agentic document workflow from scratch. Then you'll learn how to use ADE to process complex documents and build a pipeline to parse mixed documents and extract required key-value pairs. You'll also learn to integrate ADE extracted information into RAG applications. And then implement a cloud-based version on AWS using an event-driven architecture that automatically triggers ADE for document processing whenever new documents appear. Many people have worked to create this course. I'd like to thank from LandingAI, Ava Xia, and from DeepLearning.AI, Hawraa Salami and Christopher Policastro. In the first lesson, you start by looking at traditional OCR models. Let's go to the next video and dive in.