The latest articles on DEV Community by Drew Jenkins (@drewworks). https://dev.to/drewworks 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%2F4210928%2F75b3cfd0-71df-441d-b687-22402f66914e.jpg https://dev.to/drewworks en Drew Jenkins Tue, 27 Oct 2026 15:21:23 +0000 https://dev.to/drewworks/shipping-a-model-behind-fastapi-without-the-notebook-41ch https://dev.to/drewworks/shipping-a-model-behind-fastapi-without-the-notebook-41ch <p>Turned a notebook that had been "almost ready" for three weeks into an actual service this week. The model didn't change — the packaging did.</p> <p>What made the difference:</p> <ul> <li>One <code>predict.py</code> with a pure function, no global state</li> <li>A tiny test that loads the model and runs a fixed input</li> <li>Docker with a pinned wheel, not a <code>pip install</code> at boot</li> </ul> <p>The notebook is still useful for experiments. It just isn't the deployment artifact.</p> <p>Next: I want to try compiling one of these small models to WebAssembly so it can run client-side. If you've done that with a PyTorch model, I'd love to hear what worked.</p> machinelearning python api