Mark Ellis

Senior Backend Engineer · Northwind · Human-led · interviewed by Maya Chen · Jul 21, 2026 · 28 min

Senior Backend Engineer · Northwind · AI-led · Standard · interviewed by the AI interviewer · Jul 21, 2026 · 21 min

Your decision

ProceedConsiderDecline

Still your recruiter’s call.

Fit at a glance

Must-haves first · one line in the candidate’s words · each quote with its moment in the recording

Cuts API latency with measured results 9/10 Demonstrated We moved the price lookup into a cache — p95 went from 840 ms to 310 ms. at 10:12at 6:14
Keeps services covered by tests and CI 8/10 Demonstrated Every pricing change ships with a contract test in CI. at 7:26at 9:05
Owns production incidents end to end 8/10 Demonstrated …we rolled back first and only then went looking for the root cause. at 12:31

Works closely with product and design 6/10 Partial I usually joined once the designs were final. at 17:58at 15:48
Mentors other engineers 8/10 Demonstrated I paired with two new engineers through their first on-call week. at 23:40at 18:36

Draft scorehigh confidence

Drafted against your job description and your questions. Your recruiter decides.

No signals flagged

Clear

These signals do not decide the outcome.

Maya Chen

AI Interviewer

Mark Ellis

Recording · 28:14Recording · 21:06 Transcript · synced Secure Check · Clear

Maya ChenRecruiter10:05
You said you cut p95 latency on the checkout API. Walk me through one specific change you made and how you measured it.
Mark EllisCandidate10:12
We moved the price lookup into a cache — p95 went from 840 ms to 310 ms. I measured it on a week of production traffic, before and after.
Maya ChenRecruiter10:56
What was the actual bottleneck?
Mark EllisCandidate11:03
We were re-pricing the whole cart on every request. Caching the price lookups took that off the hot path.

AI InterviewerAI11:52
Tell me about a production incident you were on the hook for. What did you check first, and how did you decide when it was safe to call it resolved?
Mark EllisCandidate12:08
The first thing I did was pull up the error-rate dashboard and check whether the spike lined up with the deploy.
Mark EllisCandidate12:31
It did, so we rolled back first and only then went looking for the root cause.
AI InterviewerAI12:34
You mentioned rolling back before the error rate settled. What told you a rollback was safer than a forward fix — and how long did you wait before calling it resolved?

Draft scores are prepared from the transcript and your rubric, then reviewed by your team. A scorecard is never a hiring decision.