Reading Notes 2025 May - Jun

My Medium Articles!

In the past two months, I continued writing articles per month on Medium and Towards Data Science. Here you go:

1. The Secret Power of Data Science in Customer Support: Most data science stories focus on product or marketing. But Customer Support is another data goldmine. n my latest article, I walk through how I have worked closely with our Customer Support (CX) team, helping them track performance, plan resources, optimize internal processes, and identify customer pain points.
2. Rethinking Data Science Interviews in the Age of AI: AI is rewriting the day-to-day of data scientists. This transformation also poses a challenge to hiring managers: how to find the best talent that will thrive in the AI era? In this article, I discuss what hiring managers and candidates should do to adapt.

Reading List in Past Two Months

Now, let’s talk about the great articles I came across in the past two months.

Data Science & Analytics

1. Using Causal Inference for Measuring Marketing Impact: How BBC Studios Utilises Geo Holdouts and CausalPy: How BBC Studio uses Geo Holdout-Based Bayesian Synthetic Control to evaluate the impact of OOH campaigns.
2. Anomaly Detection in Time Series Using Statistical Analysis: Engineers at Booking.com talk about how they used statistical methods to build an anomaly detection system.
3. Statistically Speaking: How To (Properly) Report A/B Testing Results: Common errors with reporting A/B testing results, including overstating certainty, confusing test settings with test results, misinterpreting p-values, misinterpreting confidence interval, and ignoring external validity.
4. I Teach Data Viz with a Bag of Rocks: Using a bag of rocks to illustrate the principle of data visualization.
5. Time Series Linear Regression Explained: How linear regression can work for time series forecasting.
6. You can have it all: Parallel Testing in A/B Testing: Explains why running multiple experiments simultaneously is not only feasible but also beneficial, explore its key advantages and potential challenges, and share best practices for successful implementation.
7. How to Find the Right Distribution for Your Data: A Practical Guide for Non-Statistician: How the author created a visual tool for people to test distribution of their data.
8. Get More Explainability Than Just SHAP With ALIBI In Python: Alibi is a new open-sourced Python library that help with model explainability and works with both black-box and white-box explainability on local and global insights.
9. The Metric Tree Trap: A Metric Tree is a hierarchical decomposition of a top-level business goal into actionable sub-metrics. Why it could be misleading and doesn’t work as expected.
10. How to Lose Money With “Statistically Significant” Decisions: Trade-offs to consider for experimentations other than statistical significance.
11. The 10 Weirdest, Most Brilliant Algorithms Ever Devised and What They Actually Do: 10 unconventional yet brilliant algorithms—from Marching Cubes to quantum‐inspired methods—that have (or could) revolutionize fields like graphics, cryptography, optimization, and fault tolerance.
12. Compelling New Visualization Picks for Inspiration — DataViz Weekly: Five interesting new visualizations across different topics.
13. From Default Python Line Chart to Journal-Quality Infographics: A very practical step-by-step breakdown of how to turn a default matplotlib line chart to a professional-looking, clean visualization.
14. Meta’s Centralized Approach to Decision Record-Keeping: How Meta built a centralized catalog of experiment decisions to enable record-keeping and long-term decision making.
15. Holdout Groups Need Not Be a Lost Opportunity: How to determine the control group percentage under different scenarios.
16. How Did Airbnb Build Their Semantic Layer?: A great walkthrough of Airbnb’s data infrastructure evolution and the design principles of Airbnb Minerva.

Data Career

1. Should Data Scientists Pivot to AI?: The ML Researcher to AI Engineer profession spectrum and how to take a leap into AI.
2. What Data Engineers Honestly Want To Tell Data Analysts: Important data engineering related knowledge that can help data analysts to collaborate better with data engineers.
3. The Dark Side of Data Science Jobs: The reality behind the fancy data scientist title.
4. 3 Reasons Why Data Science Projects Fail: Things you should avoid to thrive as a data scientist: 1. The solution was not actionable. 2.The observational data and causal insight conundrum. 3. The solution was overly complex.

AI and LLM

1. The Biggest Problem With Text To SQL Workloads, And How To Fix It.: Challenges with TexttoSQL system come from the end user behavior and how metadata can help.
2. 99% of AI Startups Will Be Dead by 2026 — Here’s Why: Why the OpenAI API wrapper startups have a fragile foundation and can easily die.
3. Let Users Talk to Your Databases: Build a RAG-Powered SQL Assistant with Streamlit: Build a database-agnostic RAG pipeline allowing users to access data in Amazon Redshift, BigQuery, and a SQLite database.
4. What is the Future of Power BI and Business Intelligence?: Six trends of BI tools with the development of data roles and under the age of AI.
5. The BI Industry Is Missing Its ChatGPT Moment: Generative AI is starting to creep into BI but not in the way we need it to.
6. How I Automated 80% of My Data Analysis Using AI Tools: A great example of how to use AI to improve DA workflows and more ideas.
7. Stop Chasing “Efficiency AI.” The Real Value Is in “Opportunity AI.”: Opportunity AI means using artificial intelligence to solve previously impossible problems and create entirely new business and operating models. How to make your companies AI-Native?