The latest articles on DEV Community by Sam Ellis (@ellis_40). https://dev.to/ellis_40 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4210592%2F6fce03fb-6e45-4df7-85e7-31b2bd258f41.jpg https://dev.to/ellis_40 en Sam Ellis Tue, 27 Oct 2026 15:41:12 +0000 https://dev.to/ellis_40/model-versioning-is-a-data-problem-1e6b https://dev.to/ellis_40/model-versioning-is-a-data-problem-1e6b <p>We kept having the same argument: which model is in production, and why does it behave differently from staging?</p> <p>The answer was almost never the model file. It was the data.</p> <p>What we changed:</p> <ul> <li>Every training run records a hash of the input dataset.</li> <li>The serving layer logs the model version and the feature version.</li> <li>A small registry maps <code>model-version -&gt; data-hash -&gt; metrics</code>.</li> </ul> <p>That last piece is just a table, but it ended most of the "why is it different" conversations.</p> <p>The model is the easy part to version. The data is where the surprises live.</p> <p>Side quest: I'm learning Zig because I want to understand how the fast inference runtimes are built. If you've read any of them, I'd take recommendations.</p> machinelearning python data