Data Lifecycle Management as Foundational Pillar to AI Agents
summary
Claims that model accuracy falls from 99% to 74% on live data because of unqualified inputs, and that organizations often lift 10+ petabytes to cloud without attesting usability. Describes a practical shift-left workflow: profile and tag data before migration, enforce lineage, sensitivity tags and tokenization, then run cost-aware qualification so only 60%‑impact datasets move. Recommends feature stores, cataloging and expanded observability (lineage, business metadata, transformations) to keep agents reliable.
AI-written summary · about
chapters (8)
- 0:00Introduction and speaker background
- 0:44AI success challenges and data importance
- 2:55Data migration and cloud challenges
- 5:23Data lifecycle management foundation
- 7:30Common data quality challenges
- 9:14Privacy security and model governance
- 11:40Data observability and metadata principles
- 13:44Key principles for AI data foundation
related talks
16mCloud-Native Data Lifecycle Governance for Trusted AI & BI
39mLessons Learned: Insights from 30 Years in Cloud and AI/ML Integration
26mThe Hidden Foundation: How Effective Data Management Determines AI System Success
25mImportance of Data Maturity before embracing AI
27mNavigating the AI Project Lifecycle
08mLeveraging data insights to increase efficiency and innovation
30mData Prep Ops: The Missing Piece of the MLOps Puzzle
38mRedefining Test Data Strategy in the Gen AI Era