Building an Advanced Knowledge Assistant
summary
Explains a concrete stack: use a better parser (LlamaParse) to convert PDFs PowerPoints Excel into semantically coherent chunks, store text and images for multimodal RAG, and combine vector search with SQL or knowledge graphs for richer retrieval. Describes an agent layer (event driven, composable workflows like workflows that act like Airflow/Temporal) to do query decomposition, tool use and report generation. Claims production traction: 30k users and 50M+ pages processed, and shows this reduces hallucinations and enables automated report outputs.
AI-written summary · about
chapters (15)
- 0:00Introduction and agenda
- 1:04Limitations of standard RAG
- 4:20Core components of advanced assistants
- 6:46High-quality data and retrieval layer
- 8:31Advanced document parsing challenges
- 12:04Introducing LlamaParse
- 13:26Hierarchical indexing and retrieval
- 15:55Multimodal RAG pipelines
- 18:22Graph RAG and knowledge graphs
- 20:33Hybrid vector and SQL search
- 22:00LlamaCloud production platform
- 22:41Agentic reasoning and RAG
- 27:37Building agent frameworks
- 29:00Enterprise use cases for agents
- 31:04Conclusion and future directions
related talks
29mThe Future of AI Assistants
36mBuilding Agentic AI Assistants: Lessons from productionising agents in Preclinical R&D
39mAgentic RAG Platform for AI-Assisted Workflows
20mEnhancing RAG with Multi-Agent Integration
30mBuild an AI knowledge fabric for your organisation
15mGraphRAG-Powered AI Agents: Transforming Corporate Metadata into Intelligent Knowledge
29mAgentic GraphRAG: Multi-Agent Knowledge Graph Construction for Research Teams
26mAdvanced Document Parsing for AI Workflows