One question. More depth.

Implement middleware for your APIs to handle authentication and authorization. Walk through what you would check before allowing a request.

Your checkout calls a payment provider. How would you handle a request that times out?

Before creating an order, your API checks that a product is in stock. How would you implement the reservation?

Product pages are slow. You add a cache for product information, including prices.

A queue consumer processes order events and grants loyalty points.

An authenticated user requests an invoice by its ID. Which access checks should the API perform?

A frequently used database query is slow. What would you investigate before adding an index?

Limit each account to 100 API requests per minute.

Create an order in a database and publish an order-created event to a broker.

Design pagination for a feed sorted by newest first.

Your API calls three downstream services before returning a response.

Rename a database column used by an API while keeping the service available.

Your system sends writes to a primary database and reads to replicas.

Design an API for uploading large user files over unreliable connections.

Add request logging to make production API failures easier to investigate.

Build a search box that loads results as the user types.

Design a confirmation dialog for deleting a saved item.

A dashboard becomes unresponsive when it displays thousands of records.

Make a task’s completed checkbox update instantly when clicked.

Change an API response field from a string to a structured object.

Your web app uses a session cookie to authenticate requests that change account settings.

Choose the health checks and alerts for a checkout service.

A new release causes a spike in errors. What is your recovery plan?

Define a backup plan for an important database.

A checkout test fails intermittently in CI but usually passes locally.

Plan testing for a new payment integration.

A release changes login, a reporting filter, and a footer. Testing time is limited.

Engineers say the current service is hard to change and should be rewritten.

A customer commitment has a fixed launch date, but the team’s estimate is twice the available time.

Two experienced engineers disagree on an architectural choice and the team is blocked.