Trusted Agentic AI:Architecting Systems with Reliable Data at Scale
summary
Claims an AI radiology pipeline finds lesions in 10,000 abdominal images per week and raises cancer survival by earlier detection; an agentic analytics product NarraViz AI builds role-aware dashboards from spreadsheets, Snowflake, Databricks using LangChain, two LMs, a vector DB, DuckDB, Supabase and Google Cloud. User data never reaches or trains the LMs, feedback and a monitoring dashboard keep a human in the loop, and an adaptive visualization selector chooses outputs. Focus is on minimum hallucination, data privacy, metadata-driven context workflows.
AI-written summary · about
chapters (10)
- 0:00Introduction and purpose
- 3:01What is trusted agentic AI
- 5:31Key principles for trusted AI
- 7:40Demo of agentic data platform
- 10:51Technical architecture overview
- 13:26Guardrails and metadata in AI
- 16:37Traits of reliable AI systems
- 18:24Tips for AI implementation
- 20:13Six P model for AI readiness
- 22:03Summary and call to action
related talks
19mLeveraging Cassandra to Provide Generative AI That Powers Enterprise Analytics
47mThomas in't Veld | AI Track | Compass AI & Data Summit 2025
28mBeyond Dashboards: Architecting AI Agents for Autonomous Observability
29mBuilding Trust in AI Agents
46mPath of Data & AI Challenges
26mAI – Analytics made Invisible
32mGenerative AI for Observability: Automating Root Cause Analysis in Modern IT Systems
58mAI-Enhanced Big Data: Integrating Private LLMs and Vector Databases