Senior Software Engineer Interview Simulator — a staff engineer who has carried a pager
The person on the other side of this one has been paged at three in the morning and has written the postmortem afterwards. It changes the questions. They are not testing whether you know what a feature flag is; they are asking what your rollback plan was, how big the blast radius would have been, and what actually changed after the incident. Stories about architecture you designed but never operated tend to come apart here.
The kind of engineer asking the questions
Staff level, still close to the code, fluent in the difference between a root cause and a contributing factor. Blameless postmortems, backwards-compatible migrations, dual writes and backfills, API contracts and versioning, idempotency, tech debt as interest rather than a moral failing, and the long cost of a bad abstraction — these come up because your story invites them, not as a checklist. If you say latency improved they will ask p95 or p99, and whether you had an error budget to spend in the first place.
What this room actually tests
Technical disagreement resolved, including the times you were the one who was wrong. Incident ownership at the level of what you personally did, minute by minute, and what changed afterwards. How you argued tech debt against feature pressure and what you traded to win it. Code review conflict handled without ego. Mentoring measured as a concrete change in another engineer's behaviour, not as time spent. Cross-team API disputes and how the contract finally got settled. And at least one migration war story with the rollback plan, the data, and the blast radius attached.
What will get you pushed on
Hero narratives where one person saved everything — the follow-up is always "and where was the rest of the team?". Blame pointed at teammates, product, or "the requirements". "We improved performance" with no numbers, no method, and no before-and-after, which is the single most common answer this persona interrupts. And architecture described in detail that you never personally built; two questions about operational behaviour usually settle that one.
How difficulty changes this particular round
Warm-up accepts approximate numbers and lets you reconstruct the timeline out of order. Realistic asks for the measurement method behind any performance claim and expects your incident story in sequence. Nightmare assumes the outage was partly your design's fault and asks you to defend the decision that led there, then asks what you would have needed to detect it earlier — and will notice if your p95 figure moves between the two answers.
What the Sim Report grades
Four dimensions here: ownership specificity, technical precision, disagreement handling, and operational judgment. Precision is scored strictly — a claim with a method and a number beats the same claim without one, every time. Each dimension quotes the moment from your transcript and shows a tighter version of what you said. Pace and filler are reported too, since engineers commonly lose this round by narrating an entire system before answering the question asked.
How it differs from The Panel
This is the behavioral and leadership-signal round: incidents, disagreement, mentoring, migrations. It is not a system design whiteboard and it does not run coding. If you want the design round — QPS estimates, caching and invalidation, sharding, back-pressure, degradation strategy — run The Panel instead, and run both if you have a full loop coming up.
Frequently asked
›Is this a coding interview?
No. It is the behavioral and ownership round that sits alongside coding in most senior loops. Nothing is compiled and nothing is typed.
›What if I have never been on-call?
Say so early. The persona will pivot to migrations, review conflict, or debt arguments instead of pretending you have an incident story.
›How specific do the numbers need to be?
Specific enough to be checkable — a percentile, a duration, a request rate. "Significantly faster" gets a follow-up every single time.
›Does it adapt to my stack?
It follows whatever you describe. Paste your context before starting if you want the questions aimed at a particular system.
›Can I use it to prepare for a staff-level loop?
Yes — set Nightmare and describe staff-scope work. Cross-team contract disputes and blast-radius reasoning are where staff rounds separate people.
›Will it correct me if I'm technically wrong?
Not mid-answer. It will push, and the Sim Report will name the claim it did not believe and why.
Related simulators
This persona has heard "we improved performance" more times than you have said it.
Run the engineering round free