"Worked on backend services" could describe almost any software engineer's day-to-day, which is exactly the problem — it gives a recruiter nothing to search for and nothing to remember you by. Strong software resumes name the system, the scale, and the metric that moved because of the work.
Below are before/after examples across common software engineering scenarios, plus the skills worth naming explicitly. See the full ATS-friendly resume guide for the underlying bullet-writing formula.
Before / after examples
Illustrative examples — write your own numbers from your actual work; approximate is fine, invented is not.
Performance / latency
// beforeWorked on backend services and fixed bugs reported by QA.
// afterRedesigned the caching layer of a Node.js microservice, cutting p95 API latency from 800ms to 120ms and reducing infrastructure spend by an estimated 30%.
System design / architecture
// beforeHelped break down a large application into smaller services.
// afterLed the decomposition of a monolithic billing service into 4 independently deployable microservices, cutting deploy time from 45 minutes to 8 and enabling independent team ownership.
Testing / reliability
// beforeWrote tests to improve code coverage.
// afterIntroduced contract testing between 6 internal services, cutting integration-related production incidents by roughly 50% over two quarters.
Cost / infrastructure
// beforeManaged our cloud costs and monitored usage.
// afterRight-sized compute and storage across 40+ AWS services and cut monthly cloud spend by $12k with no measurable impact on latency.
FAQ
Should I list every programming language I've touched?
No — list what you're actually proficient in, grouped by category. A long undifferentiated list reads as padding and dilutes the languages that matter most for your target role.
How do I quantify impact when I don't have exact metrics?
Use a reasonable, honestly-labelled estimate — "an estimated 20% reduction" — or quantify scope instead: request volume, team size, number of services touched, users affected.