Ep214: Teradata, Amazon Bedrock AgentCore Unlock Zero-Data-Movement Analytics
Notas do episódio
Curious how AI can query your enterprise data without moving it or making things up? AWS and Teradata break down a trustworthy analyst agent built for real production use.
Topics Include:
- Neha Wadhera (AWS) introduces Trinath Yarlagadda and the Teradata Analyst Agent
- Enterprise AI data prep is costly, stalling most orgs at experimentation
- Agent answers plain-English questions via traceable SQL, zero data movement
- Barrier removal drives 3.7x ROI and 40% productivity gains
- Healthcare demo setup: hospital COPD readmissions, ~$10K cost per incident
- Four design principles: traceability, no data movement, deterministic-first, governance as code
- Main orchestrator agent plans, writes SQL, calls Teradata MCP server
- Complex questions escalate to a context-isolated data scientist agent
- Built on Claude Agent SDK, running Bedrock Claude Sonnet/Haiku/Opus
- Live demo: COPD readmission rates explored through iterative agent reasoning
- Delegation demo: data scientist agent runs in-database analysis, surfaces factors
- Pre/post tool hooks log every step and cost to CloudWatch
- Agent hosted on Amazon Bedrock AgentCore, fully serverless and scalable
- AgentCore delivers runtime, memory, identity, and observability out of the box
- Lessons learned: guardrails first, deterministic ops, multi-agent registry, ongoing evaluation
Participants:
- Trinath Yarlagadda – Principal Solution Architect – Agentic AI, Teradata
- Neha Wadhera – Sr Solutions Architect, Amazon Web Services
See how Amazon Web Services gives you the freedom to migrate, innovate, and scale your software company at https://aws.amazon.com/isv/