Product design question

Design an AWS service to help customers digitize their paper receipts.

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Technical product design for a cloud service, requiring the candidate to reason about OCR accuracy, storage, and a clear API surface.

Clarify the user: individual AWS customers building apps versus enterprises needing bulk receipt digitization.
Identify the core job: convert paper receipts into structured, searchable data reliably.
Design the pipeline: image upload, OCR extraction, structured field parsing, and confidence scoring.
Decide on the API surface: a simple endpoint returning structured JSON with vendor, date, amount, and line items.
Address edge cases: low quality scans, handwritten receipts, and non English text.
Define success with extraction accuracy rate and developer adoption measured by API call volume.

Designs a clear technical pipeline, upload, OCR, parsing, confidence scoring, rather than a vague 'use AI' answer.
Proposes a concrete API contract, structured JSON with named fields, appropriate for an AWS developer audience.
Names real edge cases, handwriting and non English text, that any OCR based service must handle.
Sets a measurable bar, extraction accuracy rate, since this is a data quality product.

Describes the product only from a consumer app perspective, missing that this is an AWS developer facing service.
Ignores the accuracy and edge case challenges inherent to OCR.

How would you handle receipts in languages the OCR model was not trained on?
What confidence threshold would you use before flagging a receipt for manual review?
How would you price this as an AWS service?