Sponsored Session: Trust by Design: Engineering Reliable Enterprise AI Workflows with Open Source
summary
An open-source reference architecture is built and demoed end-to-end: Docling and DocLang convert messy PDFs and tables into a machine-readable schema so retrieval grounds correctly, OpenRAG/OpenSearch plus a locally served 8B Granite model perform retrieval and generation, and Melia validation loop enforces deterministic requirements (catching a missing footnote that lowered a risk rating from High to Medium). Tool calls route through a ContextForge gateway that blocks unauthorized actions. The pipeline runs locally and after requirement-driven repair produces auditable, approved actions.
AI-written summary · about
chapters (12)
- 0:00Trust by Design concept
- 12:10Reliability issues in AI workflows
- 13:37Trust in workflow steps
- 15:49Docling for documents
- 19:48Docling for AI-native format
- 21:54Agent workflows with OpenLeg
- 22:38Granite model certification
- 24:07Melodia for execution
- 28:58Validating AI outputs
- 31:47Tool access with Context Bouncer
- 33:56Open-source trust engineering
- 35:40AI trust as responsibility
related talks
22mStructuring the Unstructured : Empowering the AI Workflows
40mRolling out Enterprise AI: Tools, Insights, & Team Empowerment
38mIntroducing Open Platform for Enterprise AI
38mBuild AI-Ready Search: Integrating Docling with OpenSearch for Advanced RA...
26mSponsored Session: Navigating Open Source AI with Guardrails
43mA journey from static documentation to generative AI
16mTrust by Design: Building Reliable AI for Sensitive Environments
11mKeynote: Trusted AI: Bringing Trust Back into AI through Open Source