Document AI: From OCR to Agentic Doc Extraction
by DeepLearning.AI, LandingAI
Go from classic OCR and PaddleOCR to agentic document extraction and a serverless RAG pipeline on AWS in six hands-on labs.
Overview
Document AI: From OCR to Agentic Doc Extraction is a DeepLearning.AI short course built with LandingAI and taught by David Park, Senior Director of Applied AI at LandingAI, and Andrea Kropp, Applied AI Engineer at LandingAI. At about 3 hours 11 minutes across 15 lessons and six labs, it is roughly double the length of most DeepLearning.AI short courses. It opens with document processing basics and a first OCR lab, traces four decades of OCR evolution, and has learners run PaddleOCR. It then covers layout detection and reading order, and has learners build their own agentic document-understanding workflow that combines visual layout detection with LLM reasoning. The second half moves to LandingAI's Agentic Document Extraction (ADE) API, which converts PDFs and other files into LLM-ready Markdown and JSON while preserving layout context, using visual grounding and schema-based extraction of named fields; two labs work through ADE directly. Learners then use ADE output for retrieval-augmented generation, deploy an event-driven, serverless RAG pipeline on AWS with Bedrock, and finish with a research-paper chatbot built on Strands Agents. Readers should know that the later labs centre on LandingAI's commercial ADE product, while the earlier OCR, PaddleOCR and layout-detection lessons stand on their own and transfer to any document pipeline.
At a Glance
- Topic
- RAG
- Level
- Intermediate
- Format
- Course
- Cost
- Freemium
- Duration
- ~3h10m, 15 lessons with 6 hands-on labs plus a quiz, self-paced
- Provider
- DeepLearning.AI, LandingAI
- Hands-on
- Yes — code/exercises
- Certificate
- Available
What You’ll Learn
- ✓Run a traditional OCR pipeline on real documents in a hands-on lab
- ✓Trace four decades of OCR evolution and run PaddleOCR on sample documents
- ✓Apply layout detection and reading-order analysis before passing document text to an LLM
- ✓Build an agentic document-understanding workflow combining layout detection with LLM reasoning
- ✓Parse complex files into structured Markdown and JSON with LandingAI's ADE API
- ✓Map extracted information to named fields using schema-based extraction with visual grounding
- ✓Feed agentic extraction output into a retrieval-augmented generation pipeline for question answering
- ✓Deploy an event-driven serverless RAG pipeline on AWS and a Strands Agents chatbot
Highlights
- •Six hands-on labs across 3h11m, unusually deep for a DeepLearning.AI short course
- •Covers the full arc from classic OCR and PaddleOCR through layout analysis to agentic extraction, not only one vendor API
- •Ends in a deployable architecture: event-driven serverless RAG on AWS with Bedrock and a Strands Agents chatbot
- •Taught by LandingAI's applied AI team, the builders of the Agentic Document Extraction product used in the later labs
- •Listed as a free short course by LandingAI; the graded quiz and certificate require DeepLearning.AI Pro
Who It’s For
Best For
- ✓Engineers building RAG systems over PDFs and other unstructured documents
- ✓Teams automating extraction of structured fields from business documents
- ✓Developers comparing an agentic extraction API against open-source OCR stacks such as PaddleOCR
Prerequisites
- •Basic Python familiarity
- •A working idea of what an LLM and a RAG pipeline are, for the second half
FAQ
What is Document AI: From OCR to Agentic Doc Extraction?
Document AI: From OCR to Agentic Doc Extraction is a 3h11m intermediate DeepLearning.AI short course built with LandingAI for developers who need structured data out of PDFs and other documents. You progress from traditional OCR and PaddleOCR through layout detection to LandingAI's Agentic Document Extraction API and an event-driven RAG application on AWS.
Is Document AI: From OCR to Agentic Doc Extraction free?
Document AI: From OCR to Agentic Doc Extraction offers free content, with paid options for certificates or premium features.
What level is Document AI: From OCR to Agentic Doc Extraction for?
Document AI: From OCR to Agentic Doc Extraction is aimed at a intermediate audience. Recommended background: Basic Python familiarity, A working idea of what an LLM and a RAG pipeline are, for the second half.
How long does Document AI: From OCR to Agentic Doc Extraction take?
Expect roughly ~3h10m, 15 lessons with 6 hands-on labs plus a quiz, self-paced. Most learners work through it at their own pace.
What will I learn from Document AI: From OCR to Agentic Doc Extraction?
You'll learn: Run a traditional OCR pipeline on real documents in a hands-on lab; Trace four decades of OCR evolution and run PaddleOCR on sample documents; Apply layout detection and reading-order analysis before passing document text to an LLM; Build an agentic document-understanding workflow combining layout detection with LLM reasoning; Parse complex files into structured Markdown and JSON with LandingAI's ADE API; Map extracted information to named fields using schema-based extraction with visual grounding; Feed agentic extraction output into a retrieval-augmented generation pipeline for question answering; Deploy an event-driven serverless RAG pipeline on AWS and a Strands Agents chatbot.
Topics
Sources
This page was written from 3 sources, 2 on domains other than deeplearning.ai.