Your own work
Talking about what you have built
The round engineers prepare for least, and the one most likely to decide your level rather than merely your outcome.
There is a widespread belief that this round is a formality — that it measures polish rather than substance, and that anyone competent will pass it by being pleasant for forty-five minutes. That was my assumption too, and it is wrong in a specific way: at senior levels the questions are the main evidence for scope and ownership, and scope and ownership are what levelling decisions are made of.
The mechanics are not subtle. The interviewer needs to write down that you have led something, that the outcome was partly attributable to you, and that you can work with people who disagree with you. If your answers do not contain that material, they cannot write it down, and the packet reads as a competent engineer with no evidence of scope.
Build an inventory, not a script
What worked for me was assembling six or seven stories from the last few years, written down in a few lines each, and then never rehearsing wording. Stories, not answers — the same project can serve a question about conflict, a question about failure and a question about influence, depending on which part you tell.
A reasonable spread to have covered:
- Something you led end to end, where you chose the approach.
- Something that failed, ideally through a decision of yours.
- A technical disagreement with someone whose opinion mattered.
- Something you shipped under a constraint you disagreed with.
- A time you changed your mind given new information.
- Something you did that was not assigned to you.
- An incident: what broke, what you did, what changed afterwards.
Write these as bullet points, never as prose. A memorised paragraph is audible from the first sentence and it reads as coaching rather than experience. Bullets keep the facts available and force you to actually talk.
Structure, lightly
The usual advice is STAR — situation, task, action, result. It is fine as a checklist and bad as a delivery style. What it is really encoding is that most people under-supply two things: enough context for the difficulty to be legible, and a concrete outcome at the end.
The proportions that worked:
Two or three sentences. Enough that the listener understands why this was hard. Skip the org chart and the product history.
The bulk of it, and specifically the decisions. What you chose, what you chose against, what you did when the first attempt did not work. First person singular — "we" is generous in life and useless in an interview.
One or two sentences with something measurable in them, and honesty about what remained unsolved. "Latency went from about two seconds to under three hundred milliseconds; the batch path stayed slow and we never got to it" is more credible than an unqualified success.
The uncomfortable questions
Three come up constantly and are worth thinking about in advance, because the instinct in each case is subtly wrong.
"Tell me about a conflict with a colleague."
The instinct is to pick a mild disagreement and resolve it tidily. This reads as evasion. Pick a real one, describe the other person's position in a way they would recognise as fair, and say what you did to resolve it — which sometimes means saying you were persuaded. What is being scored is whether you can hold a disagreement without turning the other person into an obstacle. A story where you were simply right and eventually vindicated is the weakest possible answer.
"Tell me about a failure."
The instinct is a disguised success — worked too hard, cared too much, shipped late because quality mattered. Everyone has heard it. Pick something that actually went wrong through a decision you made, say what you would now do differently, and stop there. The recovery matters more than the failure, and the absence of defensiveness matters more than either.
"Why are you leaving?"
The instinct is either an implausibly positive answer or an honest complaint. Neither lands. Something specific and forward-looking, without disparaging anyone, is enough — and the truthful version is nearly always available if you frame it as what you want next rather than what you are escaping.
Questions to ask
You will get five or ten minutes at the end. It is scored, lightly, and it is genuinely useful to you, which is a rare combination.
Ask things whose answers would change your decision. What does the first three months look like. How does work get chosen — assigned, or picked up. What happens when a release breaks something. How long has the team existed and how many people have left it in the last year. What is the thing about working here that people find hardest.
Avoid questions answered on the careers page, and avoid the ones that are transparently designed to sound impressive. Interviewers notice, and a candidate who is visibly evaluating the job is more convincing than one performing enthusiasm.
One small mechanical thing that helped more than it should have: take notes during the round, on paper, visibly. It slows you down in a useful way, gives you something to do while thinking, and makes the closing questions specific rather than generic.
Back to the index, or read about these notes.