CloudForestX

Filter state went global in Redux Toolkit rather than living on each page. Forty dashboards read the same account and month selection, so page-local state meant forty places to keep in sync and a filter that silently reset on navigation. The cost is boilerplate on every slice, paid once, for one source of truth.

Overlapping requests are resolved with a request-id guard rather than cancellation. Changing a filter fires several fetches and the slowest can land last, painting a chart with the wrong month. Each call takes a sequence number and only the newest result is allowed to write. It is a few lines instead of an abort-controller layer, and it cannot leave stale data on screen.

Enterprises running hundreds of AWS accounts cannot tell you what any of it costs, or which of it is doing nothing. The platform answers both questions, then recommends what to resize.

This is the engagement where I worked both ends. The ingestion path that pulls hundreds of AWS accounts into something a chart can consume was mine, which meant the API shape and the render budget were designed against each other rather than negotiated across a team boundary.