Preprocessing Unstructured Data for LLM Applications

This free short course teaches techniques for extracting and processing diverse document formats—PDFs, PowerPoints, Word documents, HTML, and EPUB—to enhance Retrieval Augmented Generation (RAG) systems. Taught by Matt Robinson, Head of Product at Unstructured, the 1-hour 12-minute curriculum covers content extraction and normalization into JSON format, metadata enrichment, document chunking strategies, document image analysis using layout detection and vision transformers, table extraction, and building functional RAG applications.

For product managers overseeing RAG-powered products, this course provides essential context on the data pipeline challenges engineering teams face when building retrieval systems. Understanding how unstructured data is preprocessed, normalized, and chunked directly informs product decisions about data source support, retrieval quality, and pipeline architecture. The course includes 8 video lessons and 5 hands-on code examples, maintaining beginner accessibility while covering technically substantive material.