Why attaching question-SQL pairs to your semantic model pushes accuracy past the baseline Why attaching question-SQL pairs to your semantic model pushes accuracy past the baseline Text-to-SQL systems have made remarkable progress. With a well-designed semantic model, modern LLMs can translate natural language questions into SQL with 70–80% accuracy. But that last 20–30% is stubborn. The system might generate syntactically correct SQL that doesn’t quite answer the question. Or it might miss subtle business logic....

How AI certification maps to data modeling challenges — and why the timing is perfect How AI certification maps to data modeling challenges — and why the timing is perfect When Anthropic launched the Claude Certified Architect (CCA) Foundations credential in March 2026, I didn’t expect it to feel so familiar. After 15 years in data architecture — designing Data Vaults, building dimensional models, and wrestling with metadata management — the exam domains read like...

How to collect, organize, document, and deploy specialized knowledge for Claude and other AI coding agents How to collect, organize, document, and deploy specialized knowledge for Claude and other AI coding agents AI coding agents are powerful, but they’re generalists. They’ve been trained on vast amounts of code, but that breadth comes at a cost — they often produce “plausible but wrong” output when working in specialized domains. Enter Agent Skills : structured knowledge bundles...

Make your semantic model AI-ready with systematic checks Make your semantic model AI-ready with systematic checks Microsoft Copilot in Power BI is only as good as the semantic model it works with. Feed it a poorly documented model with cryptic names and implicit measures, and you’ll get hallucinated DAX and frustrated users. This article provides a comprehensive checklist of 15 requirements for Copilot-optimized semantic models — and shows how to validate them automatically. When Copilot...

Tools, not just chat: How Model Context Protocol transforms AI assistance Tools, not just chat: How Model Context Protocol transforms AI assistance Series: Building a 6,600+ Test Framework with Claude Code Part 1: The Claude-First Architecture Part 2: MCP Integration — Tools, Not Just Chat (this article) Part 3: Practical Tips for AI-Assisted Development In Part 1, we built the documentation foundation: CLAUDE.md, ADRs, and self-documenting schemas. That’s table stakes for effective AI collaboration. Now...

How documentation-as-code transforms AI from assistant to collaborator How documentation-as-code transforms AI from assistant to collaborator Series: Building a 6,600+ Test Framework with Claude Code Part 1: The Claude-First Architecture (this article) Part 2: MCP Integration — Tools, Not Just Chat Part 3: Practical Tips for AI-Assisted Development When people ask me about AI-assisted development, I don’t show them a chatbot writing a todo app. I show them MDDE — a metadata-driven data engineering framework...

How 7,450+ AI-generated tests actually work — and what it means for test managers How 7,450+ AI-generated tests actually work — and what it means for test managers Marc Teunissen, a former colleague, asked me a pointed question after seeing that our metadata framework had over 7,000 unit tests: “How does AI generate 7,000 tests? What are they based on? And how do you reproduce bugs when the ‘developer’ is an AI?” These are exactly...

The Problem Every few months, someone claims: “AI will generate your entire data platform.” It won’t. Large language models do not understand your architecture. They do not understand lineage. They do not understand semantic stability. They generate text. Text is not structure. And SQL is not just text. AI becomes interesting only when it operates on structured metadata. Once SQL has been parsed into: entity attribute relationship attribute_mapping function_def filter_def entity_window optimizer_diagnostics The system is...

Lifecycle Traceability Many frameworks fail at the same point: They confuse detection with enforcement. They detect a pattern… …and immediately try to fix it. That works in greenfield projects. It fails in production systems. MDDE takes a different stance: Observe first. Propose second. Act only when approved. This separation is deliberate. In real systems: Queries contain historical edge cases Performance optimizations exist for specific reasons Workarounds hide migration constraints Naming inconsistencies have downstream dependencies If...

How to integrate conversational AI into your metadata-driven workflow After turning metadata into a knowledge graph and enabling semantic reasoning, the next logical step is to make it conversational. Imagine being able to ask: “Which logical entities are missing lineage?”  “Generate a YAML model for the Finance domain.”  “Summarize the difference between Customer and Client models.”  “Add a quality rule for all attributes tagged as PII.” That’s the vision of the Metadata Copilot — an AI assistant...