Databricks: AI-Driven Document Parsing and Agentic Workflows

​Format: Presentation, Demo, and Q&A

​Session Overview:

​This session will explore AI-driven document parsing, layout analysis, knowledge graphs, and agentic workflows within Databricks' open AI ecosystem. Attendees will gain insights into how compound AI systems can be used to automate document understanding and policy analysis, with a specific focus on legislative documents from the Government of Ontario.

​Key Topics Covered:

​🔹 GenAI & Agentic AI Presentation & Demo (15 mins)

• ​Overview of Databricks' open approach to compound AI systems

• ​How Databricks aligns with AI ecosystems like LogicStudio, OpenAI, etc.

​🔹 Background of Gov of Ontario Demo: Parsing & Knowledge Graphs (15 mins)

• ​Understanding document parsing, layout analysis, hierarchy, and knowledge graphs

• ​Non-technical breakdown of AI components: how agents process and evaluate data

• ​Data quality, auditing, and chatbot validation before deployment

• ​Steps of retrieving and generating context within a chatbot framework

• ​GitHub resource for multimodal AI accelerator:
  https://github.com/ScottHMcKean/multimodal-accelerator

​🔹 Live Demo: Gov of Ontario Parsing Legislation (10-15 mins)

• ​Practical walkthrough of AI-driven legal text analysis

• ​Answering legislative queries like:

  • ​“What is legislation for starting a milk farm?”

  • ​“What is legislation for _____?”