Cloudizero
Media asset optimization as a product. Smaller files, same visual quality, one API call.
Heavy images and video are the most common reason a fast site turns slow. Teams know it, but optimizing by hand does not scale, and existing tools were either too manual or too opaque.
Developers frequently pushed raw 4MB+ banner images directly to production, causing severe Largest Contentful Paint (LCP) regressions and burning through client bandwidth.
Uploads land in object storage via short-lived presigned URLs. As soon as an asset lands, jobs are queued on PostgreSQL with guaranteed processing states.
Workers pick the optimal format and quality per asset, dynamic resizing matrices are calculated, and the client dashboard reads real-time status through TanStack Query without polling latency.
Keeping quality visually identical while shrinking the file. Static compression flags either cause noticeable artifacts on intricate illustrations or waste bandwidth on flat surfaces.
I settled on a per-asset quality search driven by a perceptual error threshold, instead of a rigid global setting.