Mehdio — practical data and AI education by Mehdi Ouazza
|About Mehdi Ouazza and this personal website
Mehdio is the personal website of Mehdi Ouazza, a data engineer, developer advocate, speaker, and educator based in Brussels, Belgium. Mehdi publishes tutorials, videos, talks, open-source projects, experiments, and notes about data engineering, artificial intelligence, DuckDB, databases, and developer tools.
Explore Mehdi's work
- Blog articles and practical tutorials about data and AI
- Videos, conference talks, and live demonstrations
- Open-source projects and data tooling experiments
- About, contact, and professional background information
Agent-readable resources
The site publishes Markdown representations, llms.txt, llms-full.txt, an XML sitemap, and an RSS feed so automated tools can discover and understand its public content.
Fresh off the press
Why DuckDB 2.0 is faster
Why DuckDB 2.0 is faster
If you are doing any analysis with PyPI data (e.g. downloads of a given Python library typically), you probably want to revisit how you are calculating things because you may have seen an inevitable drop in your downloads (rest assured). PyPI published that, since August 24, they decided to stop taking into account metadata sidecars and other package objects in download logs. They now only count actual distribution artifacts: .whl, .tar.gz, and .zip. Which means that the easiest fix is to adapt your historical data too, filtering it with the same rule, so it stays coherent with future metrics. https://blog.pypi.org/posts/2026-08-31-download-counts/
If you are doing any analysis with PyPI data (e.g. downloads of a given Python library typically), you probably want to revisit how you are calculating things because you may have seen an inevitable drop in your downloads (rest assured). PyPI published that, since August 24, https://t.co/GU7Dd7SbLu