Data Quality Framework for Machine Learning and AI Agents
Bad data kills most AI projects before models even matter.
Yuki Novak
Contributing Editor
Yuki Novak
Yuki Novak covers data freshness, semantic layer for ai and ai-ready data layer for AI-ready data.
12 stories
Data Observability vs Data Quality for AI Pipelines
Most AI pipelines fail on data quality, not algorithms, and the difference costs millions.
Yuki Novak
Semantic Layer vs Vector Database for AI Querying
Choosing between them determines whether your AI gets reliable answers or confident wrong numbers.
Yuki Novak
Structural Metadata vs Semantic Metadata for AI Consumers
AI agents can't fill in the meaning that human analysts guess from context.
Yuki Novak
Descriptive Metadata as AI Agent Context
Agents need accurate, current metadata to reason reliably, not just clean data.
Yuki Novak
OLTP vs OLAP for AI Agent Data Sources
AI agents need fresh transactional data and historical patterns simultaneously.
Yuki Novak
Revoking AI Agent Data Access at Workload Granularity
Scoped credentials and revocation per task eliminate the security cost of denying standing access.
Yuki Novak
Federated Data Access for AI Workloads
Keeping data where it lives requires semantic layers to prevent silent failures in AI agents.
Yuki Novak
Institutional Memory Gap in AI Agents
AI agents repeat inconsistent answers because enterprises lack documented data context.
Yuki Novak
AI Agent Read vs Write Permission Asymmetry
Agents are provisioned with ten times the access they actually need.
Yuki Novak
Service Account Risks in Agentic Pipelines
Reusing service accounts for agents masks overpermissioned credentials behind opaque audit trails.
Yuki Novak
Knowledge Graph Integration for AI Data Retrieval
Knowledge graphs make AI retrieval factually reliable instead of semantically approximate.
Yuki Novak