How to write it
The four numbers that make a DevOps resume
Deployment frequency, lead time for change, change failure rate, and mean time to recovery. These are the metrics the discipline organised itself around, and a resume carrying two or three of them with real before-and-after values is immediately more credible than one carrying twenty tool names.
"Took release cadence from weekly to 30 deploys a day and lead time from 6 days to 45 minutes" tells a reader what kind of engineer you are, what state you found things in, and what you're capable of changing. No tool list does that.
- Deployment frequency, before and after
- Lead time from commit to production
- Change failure rate, or failed releases per hundred
- MTTR, and the alerting work behind the improvement
- Availability against the SLO you were accountable for
Cost is a first-class outcome
Cloud spend is one of the largest controllable line items in a modern engineering budget, and an engineer who has cut it materially is making an argument in a language finance understands. State it in currency per month with the percentage beside it: "$47k/month (31%)".
Say how, briefly, because the how is what proves it was engineering rather than a spreadsheet exercise — rightsizing, spot capacity, storage tiering, deleting what nobody owned. A cost number with no mechanism reads as someone else's saving.
Infrastructure as code, with the estate size
"Experience with Terraform" is the weakest possible version of a strong claim. The strong version has scale and a starting condition in it: 900 resources codified, half of which were previously hand-built, with drift detection added.
That sentence tells a hiring manager you've done the unglamorous part — importing existing infrastructure into code, which is where the real difficulty lives — rather than only writing greenfield modules.
Developer experience is the modern platform pitch
Platform teams increasingly justify themselves by what they remove from other engineers' days. Environment provisioning that used to take three days and now takes twenty minutes; a Helm library that standardised deploys across six teams; a pipeline nobody has to think about.
This framing also travels well into platform engineering roles, which are largely DevOps roles with an internal-product mindset attached. If you have built anything self-service, say who uses it and what it replaced.
Certifications actually count here
Infrastructure is one of the few areas where certifications carry real screening weight — CKA, the AWS professional tiers, Terraform Associate. Recruiters filter on them and enterprise clients sometimes require them, so list them with the issuer and the year.
They are still not a substitute for the metrics. A CKA with no reliability numbers on the page reads as someone who has studied Kubernetes rather than run it in anger.