AI-Assisted Receipt Upload and Transaction Matching

Improve an existing receipt-upload UI so users can upload receipts, observe asynchronous OCR processing, inspect extracted merchant/date/amount data, and confirm or override the suggested transaction match. Requirements Work inside an existing product codebase and improve the receipt-upload and matching experience. The existing flow makes users locate a transaction, open it, and upload one receipt at a time. A new OCR service can process an uploaded receipt image in a few seconds, extract its amount, merchant, and date, and suggest a matching transaction. Build an experience that lets the user: upload receipts; see the processing state while OCR runs; inspect the extracted amount, merchant, and date; and confirm the suggested transaction or choose a different match. Notes This is an AI-enabled interview in an existing codebase, so implementation speed and the ability to navigate unfamiliar code are part of the exercise. The OCR call is asynchronous and takes a few seconds; the UI must make that intermediate state legible instead of appearing frozen. Matching is suggested rather than final. Preserve an explicit user confirmation and correction path. Preparation Build a small practice flow end-to-end: multi-file upload, a mocked extraction call that resolves after a few seconds, per-receipt states (uploading → processing → matched / needs review), and a confirm-or-override control for the suggested match. Drill keeping the pending state visible without blocking further uploads. Rehearse AI-assisted feature work in an unfamiliar repository: while the assistant generates, read the surrounding components and API layer, review every proposed diff before accepting it, and budget time to test and explain the final behavior.