35 min
of real design in a 45-minute slot
Five minutes of introductions and five for your questions leave about 35 to turn a one-line prompt into a design.
Is there one right answer?
Bad: Find the right design
Memorise the architecture and draw it fast.
Good: Show the reasoning
No single answer, but there are wrong ones, such as designs that miss a stated requirement.
Interviewers collect evidence on five signals.
Navigating the problem, designing a solution, depth and trade-offs, scale and failure, and communication. Each part of the round feeds some of them.
The interviewer won’t remember the session, and a committee may read only the notes.
So
If you didn’t say it, it isn’t in the notes. Narrate: “X over Y because of Z; it costs W.”
The prompt is the same at every level. Who drives changes.
Mid-level reaches a working design with some steering. Senior sets the agenda and picks deep dives unprompted. Staff frames the problem and how it evolves.
Same food-delivery prompt, two very different sessions.
The weak one names Kafka by minute 6 and never gets deep. The strong one asks first, works end to end by minute 12, then finds the hard part.
An eight-week plan
- Weeks 1–2: fundamentals, plus daily estimation
- Weeks 2–5: building blocks, then patterns
- Weeks 4–8: about 20 designs, four a week
- Weeks 5–8: one or two timed mocks a week
Remember
About 35 minutes of thinking out loud, scored on five signals. Pin down requirements, reach a working design, go deep with honest trade-offs, cover failure, and say each decision with its cost.