Structuring the Unstructured : Empowering the AI Workflows
summary
Claims a practical, open-source pipeline that turns messy docs into a unified semantic model and knowledge graph, runs on CPU (Mac M2/M3 or Xeon) and supports on‑prem compliance, RAG/PAL and integrations with LlamaIndex/LangChain and Dockling’s tableformer and visionformer. Benchmarks claim roughly 1.3 seconds per PDF on CPU and about 90–95% accuracy on common tasks, with JSON/markdown exports for agent-to-agent and human-readable flows.
AI-written summary · about
chapters (11)
- 0:00Introduction to unstructured data challenge
- 2:49Why unstructured data matters
- 5:18Challenges with unstructured data
- 6:00Introducing Dockling tool
- 10:54How Dockling processes documents
- 13:21Building knowledge graphs
- 15:52Dockling output formats and integrations
- 17:59Data quality and grounding in Dockling
- 19:03Generative AI and Dockling
- 20:49Performance metrics and demo
- 21:09Q&A and closing remarks
related talks
21mStructuring the Unstructured
01h 00mComplex Data Ingestion with Open Source AI
26mAdvanced Document Parsing for AI Workflows
01h 19mLearn to Unlock Document Intelligence with Open-Source AI
30mFrom Unstructured Docs to Graph Intelligence: A Framework for Building Graph+Embedding Pipelines
38mBuild AI-Ready Search: Integrating Docling with OpenSearch for Advanced RA...
37mSponsored Session: Trust by Design: Engineering Reliable Enterprise AI Workflows with Open Source
31mThe Fast and the Curious: Harnessing AI in Financial Data Extraction at Scale