<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Alex Tarverdy — engineering writing</title><link>https://alex-tarverdy.xyz/writing/</link><description>Articles and notes on data engineering, Scala and reliable systems.</description><language>en</language><atom:link href="https://alex-tarverdy.xyz/feed.xml" rel="self" type="application/rss+xml"/><item><title>A pipeline can be green and still be wrong.</title><link>https://alex-tarverdy.xyz/writing/green-pipeline-wrong-data/</link><guid isPermaLink="true">https://alex-tarverdy.xyz/writing/green-pipeline-wrong-data/</guid><description>Why successful execution is only one part of trusting a data pipeline.</description><category>data-correctness</category><category>spark</category><category>sql</category><category>financial-data</category><category>reliability</category></item><item><title>Retries need a destination that can handle them.</title><link>https://alex-tarverdy.xyz/writing/retries-and-idempotency/</link><guid isPermaLink="true">https://alex-tarverdy.xyz/writing/retries-and-idempotency/</guid><description>What retries and primary keys can guarantee—and where the boundaries matter.</description><category>kafka</category><category>streaming</category><category>idempotency</category><category>reliability</category></item></channel></rss>