Yu Dong

Data Science & Analytics, Data Storyteller

3 minute read

Reading List in Past Two Months

Here are the great articles I came across in the past two months.

Data Science, Engineering & Analytics

  1. The 12 Data Architecture Patterns Every Data Engineer Should Master in 2026: Common data architecture patterns with explanation and use cases.
  2. The Evolution of Data Teams: How We Moved from ETL to Analytics Engineering: Companies first struggled to collect data (before 2010), then to process data (2010-2020), and now to trust data (2020 onwards) – that’s why we need Analytics Engineers.
  3. The Character’s Journey: Designing Content-Specific Analytics for Interactive 3D Environments: Meta team shares what they learned about designing analytics for interactive 3D environments, including the data architecture behind it and the customer-facing analytics product.
  4. The Medallion Architecture Reconsidered: What It Solved, and Where It’s Cracking: Why the medallion architecture — Bronze, Silver, and Gold — mihgt not fit the analytical use cases in the AI era.
  5. What an Ontology Actually Is: An ontology is an explicit, machine-readable specification of what exists and what the rules are.
  6. The Bronze-Silver-Gold Pattern Is Not Enough: Designing Enterprise Data Products in Databricks: The Medallion architecture is one of the most recognizable Databricks design patterns, but what else is necessary.

Data Career

  1. Competencies for Analytics Leadership to Manage Intelligence: 2 additional competencies that need to be mastered by Analytics leaders to succeed in managing intelligence — Systems Thinking and Data Strategy.
  2. AI Can Write the Code. These 10 Skills Keep Data Professionals Irreplaceable.: AI can write SQL faster than most of us, but it cannot tell you whether the answer is actually right, whether you’re measuring the thing the business cares about, whether a trend is real or a quirk in the data, or whether the question you were asked is even the question you should be answering.
  3. Our Data Was Clean. Our Decisions Were Still Wrong. Here’s Why.: The dangerous failures live in data that’s technically accurate and still tells you the wrong story, because of what’s missing, how it’s aggregated, or what changed since it was collected.

AI and LLM

  1. Data Product Standards Are Moving From Metadata to AI-Ready Context: Data products need a common language because organizations need data to move across teams, platforms, partners, catalogs, marketplaces, and AI workflows.
  2. How We Built DEmate: Taming LLMs for Data Engineering at Meta: DEmate, a specialized AI assistant that doesn’t just write code, but understands the unique topography of Meta’s data engineering ecosystem.
  3. 6 Data Modeling Mistakes That Trip Up Beginners (Including Me): A good data model starts with the questions the business needs to answer — not with the tables you want to create.
  4. How To Build A Reliable AI-Based Structured Data Extraction Pipeline: A reliable LLM-based extraction system should not only handle the diversity in document formats, but also the differences in organizational requirements. This article walks through an LLM-based extraction pipeline setup with Unstract step by step.
  5. Where Agentic AI Actually Helps in Data Engineering: The agent was very good at repetitive work — inspecting context, generating SQL, drafting documentation, producing initial tests. What it was not good at was independently knowing whether a business rule had been interpreted correctly.
  6. Your Data Engineering Agent Shouldn’t Be Allowed to Touch Production: The author stated why they believe we should not give data engineering agent unrestricted production access, no matter how good the demo looks.
  7. **Anthropic is telling you that Agentic Analytics is not just text-to-SQL: Anthropic’s own architecture and their stack puts data foundations, sources of truth, skills, and validation ahead of the moment when Claude ever touches a query.
  8. Announcing OpenMetadata 2.0: The Open Context Layer for AI Agents: The author walks through OpenMetadata 2.0, the next generation of their Open Context Layer for AI agents and data teams.
  9. The Semantic Layer, Explained: The Meaning Layer Your AI Agent Needs: A practitioner’s guide to what a semantic layer is, how it works, and why it’s only one of three layers a production analytics agent needs.
  10. AI Data Management: What It Actually Takes to Trust an AI Agent: AI data management is the practice of structuring, governing, and packaging enterprise data so AI systems can consume it reliably at scale.
  11. What Is an Analytics Agent and How AI Agents Are Replacing Dashboards: What analytics agents actually are, how they differ from copilots and chatbots, and why most of them fail before production.
  12. Top 7 Signs Your Data Isn’t Ready for AI: How to tell if your data is AI-ready from data infra, semantic layer, lineage, access, etc.
  13. Are Dashboards Still Relevant in the Age of AI?: A good dashboard is designed to trigger questions rather than a “collection of answers”, and it is still very useful in the AI era.

You May Also Enjoy