• The Engineering Challenge: Browser-based image utilities frequently freeze when processing raw multi-megabyte payloads and suffer from canvas distortion (stretched aspect ratios) upon re-rendering processed masks. • Client-Side Preprocessing & Memory Control: Built a strict canvas dimension lock that inspects image metadata prior to dispatch. Processing cycles operate on memory-isolated byte buffers with immediate garbage-collection triggers after export, preventing browser tab RAM ballooning. • Asynchronous API Pipeline: Engineered a non-blocking multipart streaming pipeline communicating with AI segmentation endpoints. The handler captures raw image streams, verifies dimensions against the original bounding box, and renders the transparent alpha channel without layout shift. • Production Impact & Result: 100% aspect ratio fidelity across diverse resolutions, sub-3s end-to-end processing turnaround, and zero memory leaks across consecutive multi-image sessions.