Yu Dong

Data Science & Analytics, Data Storyteller

3 minute read

My Medium Articles!

Let me start with my new article. I am finally back to writing last month:

  1. AI Made Data Scientists Faster. Now It’s Expanding the Job.: A year ago, I wrote about how AI was changing the day-to-day work of data scientists. At the time, I thought the biggest impact would be straightforward: eliminate low-value work and help us move faster on high-value analysis. A year later, I think I underestimated the change. AI has certainly made data scientists faster—but more importantly, it has expanded the job itself. We can now own more of the end-to-end workflow, take on a broader range of technical work, and delegate more execution to agents. At the same time, judgment, governance, and prioritization are becoming more important, while traditional paths for developing data science skills are being challenged. In this follow-up, I share what has changed in my own work over the past year, what surprised me, and what I think it means for the future of data science.

Reading List in Past Two Months

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

Data Science, Engineering & Analytics

  1. I Tested Google’s TabFM Against XGBoost and AutoGluon: Google’s TabFM almost removes task-specific training from the workflow. A detailed comparison betwee TabFM and XGBoost on different use cases.
  2. Ontology vs Semantic Layer: Do you need both?: The semantic layer governs how the business measures the world. The ontology describes the world being measured.
  3. From Gold Tables to Data Products: Redefining the Medallion Contract: The boundary between Silver and Gold is ambiguous. A best way to think about is that the Gold layer should also be your data contract with your consumers.
  4. Delta Table or Materialized View? Here’s how I actually decide.: Large aggregation refreshed daily from a stable source — MV wins. If CDF is enabled, incremental refresh means only new rows since the last run get processed. Complex transformation with Python UDFs, custom logic, or a streaming source — Delta table, full stop.
  5. The Semantic Layer Race: Snowflake, Databricks, Fabric IQ, and Apache Ossie: How different products are incorperating semantic layers and how you should get started.
  6. Causal Machine Learning: Moving Beyond Prediction to Real Decisions: A detailed walkthrough of causal ML, common methods, with real code example and pitfalls.
  7. The dbt Debate: When a transformation framework still adds value — and when the platform is enough.
  8. Are Tabular Foundation Models Ready to Replace Gradient Boosting Models?: An introduction of tablular foundation models and compare it with the most common gradient boosting models.
  9. We Set Out to Replace Tableau. Four Challenges Later, We Stayed.: A real story of Databricks AI/BI Dashboards and Genie vs. Tableau.

Data Career

  1. Teaching Data Analytics in the Age of AI: Data analytics space keeps changing, but core principles are still the same.
  2. How I Actually Use Claude Code as a Data Scientist (Real Workflow That Saves Hours Every Week): The author’s key learnings include context is everything, documenting assumptions aggresively, turning repeated work into skills, using Claude as a thought partner, and letting a new session reviewing the work.
  3. A New Golden Age of Analytics: Analytics isn’t about coding. It’s much more of a mental affair. And AI has interestingly better enabled those who are cognitively stronger.
  4. How Will My Career Change in the Age of AI: Careers will be increasingly non-linear in the Age of AI.

AI and LLM

  1. AI can build a Dashboard in minutes. Why Is your data team still taking Weeks?: What the faster dashboarding means for data dashboards and the data team.
  2. Scaling AI-Native Analytics Patterns at Meta: Meta team talks about how they find recurring patterns in product analytics and use AI to scale them.
  3. Claude killed my north-star dashboard: AI’s impact on data dashboards and why a great dashboard is still incredibly useful.
  4. Cortex Sense and the Shifting Role of Data Science: The snowflake team explains how they built and measured the accuracy of the newly-launched Cortex Sense.
  5. 10 AI Tools Everyone Will Be Using in Six Months, But Almost Nobody’s Found Yet: A list of AI tools to explore.
  6. Claude Dashboards vs Power BI vs Tableau: Is AI Replacing Traditional BI?: Where traditional BI tools maintain dominance and how the industry is heading into a hybrid integration.
  7. What the Research Actually Says About Semantic Layers and Text-to-SQL Accuracy: A very comprehensive overview of existing semantic layer and text-to-SQL accuracy.

You May Also Enjoy