World's Fair 2026
Structuring the Unstructured - Cedric Clyburn, Red Hat
Overview
This talk addresses the challenge of processing unstructured data, such as PDFs, presentations, and technical documents, to make it usable for AI applications. The core thesis is that effectively structuring this data is crucial for building robust AI systems like RAG and agents, and that open-source tools can provide efficient, local solutions. The presentation highlights the limitations of basic parsing methods and introduces Docling as a comprehensive tool for extracting and transforming various document formats.
Who should watch
- AI Engineers working with diverse data sources.
- Product Managers seeking to leverage unstructured data for AI features.
- Builders looking for open-source solutions for document processing.
- Developers dealing with challenges in RAG and agent development due to data format issues.
- Those concerned about data privacy and the cost of proprietary solutions.
Key takeaways
- A significant portion of the world's data is unstructured, posing a challenge for AI applications.
- Basic PDF parsers often fail to accurately extract information from complex documents containing tables, images, and varied layouts.
- Docling is an open-source, local command-line tool and library that can convert various document types (PDFs, images, websites) into structured formats like Markdown and JSON.
- It utilizes OCR and vision models to handle complex elements like tables and images, enabling structured output and enriched descriptions.
- Docling supports advanced use cases such as agentic RAG without traditional chunking and vector databases, and can be deployed as a scalable REST API service.
- The tool integrates with AI agent frameworks via the Model Context Protocol (MCP), allowing agents to leverage Docling's capabilities for document processing.
- Using Docling can lead to significant cost savings compared to proprietary solutions and VLMs, especially when processing large volumes of documents.
- The tool preserves document layout and structure, ensuring consistency and accuracy for downstream AI tasks.
Notable quotes
*Context is the most important aspect to building an AI application or an agent.*
*Unstructured data is becoming this new context layer for AI.*
*The data and the way you process it is the key determining factor in whether your answer is going to be correct or incorrect.*
Unofficial community note. Prefer the recording for nuance.