For an edtech product that stores AI-generated images, the important choice is not just how to upload bytes. It is who owns the retention decision after the upload. A presigned URL keeps access keys on the server, while a tenant-scoped object key and an explicit deletion policy keep one school's export away from another school's data.
Here is the practical choice matrix:
Option Best fit Boundary to accept
Amazon S3 A team already standardized on AWS storage and controls the surrounding AWS account You own the service integration and its storage policy surface
Cloudflare R2 An S3-oriented workflow that already fits the Cloudflare environment You still need to design tenant naming, retention, and export authorization
Alibaba OSS A deployment already centered on Alibaba Cloud storage The application must carry the provider-specific operational boundary
Infrai storage A small Node.js service that wants one discoverable HTTP surface for the upload handoff It is not a public image host, immutable archive, or…