Technique that avoids restarting large uploads from scratch when a single HTTP request fails. The file is split into chunks uploaded individually; failed chunks can be retried and multiple chunks can potentially be uploaded in parallel, improving speed. Critical for users with flaky or slow connections for whom a single request may never last long enough. Two common implementations: (1) S3 multipart upload API driven from the browser (storage-coupled), and (2) the generic tus HTTP protocol with many client/server implementations. Uppy stores upload state in browser local storage so uploads can resume across page reloads.