Amazon

What they grade on

• Operational rigor — how would you run this on-call?
• Explicit failure reasoning — what's your blast radius, what's your RCA story?
• Cost awareness — Amazon interviewers ask about $ per request
• Leadership Principles woven in — 'customer obsession' and 'ownership' should show up naturally
• Depth on storage: DynamoDB, S3 patterns come up often

Deal breakers

• No thought given to on-call or observability
• Ignoring cost implications when asked
• No answer for 'how do you know it's broken?'

Super helpful signals

• Referring to a real AWS service you'd use and its known limits (DynamoDB partition limits, S3 request rates)
• Naming an SLI/SLO you'd track
• Discussing multi-AZ vs multi-region trade-offs explicitly

Delivery advice

• State a customer scenario in the first minute. 'A customer with 1M items needs sub-100ms search' hits the customer-obsession note.
• Say what you'd track ('p99 latency', 'error budget of 0.1%') and how you'd alert.
• Volunteer a cost estimate even if rough. 'DynamoDB provisioned + on-demand switch above 20K RPS' shows AWS depth.
• Own your decisions — say 'I own that trade-off; here's why.'
• Discuss blast radius: 'if this region fails, what happens to the customer?'

Red flags to avoid

• 'That's an ops problem' — Amazon expects the designer to own operations
• Ignoring cost even when asked
• No SLI/SLO named