Bartley Richardson

AI Summary

→ WHAT IT COVERS Bartley Richardson explains agentic AI as next-level automation for enterprises, covering NVIDIA's technology stack including NEMA Retriever for multimodal data ingestion, reasoning models, AgentIQ observability platform, and context-based security approaches for distributed agent systems. → KEY INSIGHTS - **NEMA Retriever Performance:** Processes complex PDFs at 10 pages per second on single GPU with 15x throughput improvement over competitors and 50% fewer accuracy errors, handling multimodal documents with text, tables, charts while preserving contextual relationships between elements. - **Agent Ops Tools Efficiency:** Fine-tuning through successive iterations and tool emulation delivers 10x model size reduction while increasing accuracy by 4%, with human feedback loops using thumbs up/down plus free-form text to steer model behavior toward specific enterprise use cases. - **AgentIQ Observability Platform:** Reduces everything to function calls enabling cross-framework traceability across LangChain, CrewAI and other frameworks, allowing developers to inspect input/output tokens, timing, and sequences—customers achieve 15x speed improvements and 5x accuracy gains through optimization. - **Context-Based Security Model:** Moves beyond firewall and application-based security to analyze the context of each query, examining who asks, what information accompanies the request, and what data should be returned—adding 10% new security requirements to existing 90% application security practices. → NOTABLE MOMENT Richardson demonstrates how reasoning models facilitate human-in-the-loop workflows where AI generates brainstorming questions from customer issues, humans meet to discuss while drawing diagrams, then the system produces 75-80% complete product requirement documents from meeting transcripts and whiteboard photos. 💼 SPONSORS None detected 🏷️ Agentic AI, Enterprise AI Infrastructure, AI Observability, Reasoning Models