Resume coaching for software engineers: impact bullets without inflation

Engineering résumés fail in two opposite ways: listing technologies with no evidence of impact, or claiming impact in numbers that collapse under one follow-up question. The requirement side has the same subject trap as every field — "5 years building distributed systems in Go" is a language-plus-domain subject, not a total-years question — and the AI-rewriting era added a third failure: tools that inflate your latency and scale numbers into claims you can't defend in a system-design round.

Reading the requirements

• Split language requirements into quantity + subject: production years in that language, not years since you first touched it.
• "Scale" qualifiers (high-throughput, low-latency, on-call for production) are subjects — a claim of distributed-systems experience should come with QPS, data volume, or SLO numbers you personally operated under.
• Seniority words decode to scope: senior ≈ owned a system and its failures; staff ≈ owned a problem across teams. Match your bullets' scope to the level you're applying for.
• The trailing tool list is usually negotiable; the architecture verbs (design, operate, migrate, debug) are not.

Impact bullets that survive cross-examination

The convert pattern is system + your change + measured effect: "rewrote the ingestion path's batching layer, cutting p99 from 900ms to 210ms at 40k events/s". Every number there is a follow-up question — where did 900ms come from, why batching, what broke — and that's the point: engineering interviews are cross-examinations of the résumé. Which is why AI inflation is deadliest for engineers; in our logs, models invent and round numbers confidently, and a fabricated p99 gets found in the first five minutes of a system-design round. Verify every figure an AI touched against something you can re-derive on a whiteboard.

Before the interview

Rehearse the second-order questions on your own bullets: the why-this-design and what-failed follow-ups. OfferCoach builds its mock interviews from your résumé and the target JD — including a voice mode that probes the weak spots its diagnosis found — and its rewrite gate strips any number it can't verify from your history and asks you instead. Whatever you use, the standard is the same: nothing on the page you can't defend at a whiteboard.