I Screened 23 SRE Candidates in Dubai Over 6 Weeks — the 7-Step Filter That Cut Our Interview Time in Half

We opened the role because a payment integration went down for 47 minutes on a Thursday evening and nobody could say, afterwards, whether that was unusual. Not “nobody knew the cause” — we found the cause. Nobody could say whether 47 minutes was acceptable. That is the gap an SRE fills, and it is a different gap from the one most job descriptions describe.

What follows is the method that came out of six weeks and 23 candidates. It is reproducible, it is specific to the Dubai market, and it assumes you are hiring your first or second SRE rather than growing an established team.

Step 1 — Decide whether you actually need an SRE or a DevOps engineer

Do this before you write a word of the job description, because getting it wrong wastes the entire cycle and costs you a hire nine months later.

A DevOps engineer optimises how code reaches production. Pipelines, infrastructure as code, environment parity, developer velocity. The measure of success is deployment frequency and lead time.

An SRE optimises how production behaves once the code is there. Service level objectives, error budgets, incident response, capacity planning, toil reduction. The measure of success is that you can answer “was 47 minutes acceptable?” with a number rather than a shrug.

The cleanest test is authority. An SRE role that cannot say no to a release is not an SRE role. It is a DevOps role with a more expensive title. This matters commercially in Dubai because SRE-titled candidates command a 10 to 20% premium, and hiring one into a pipeline-maintenance mandate produces a predictable resignation within nine months.

Step 2 — Write a scope-honest job description

Three things belong in the description that almost never appear, and each of them filters out the wrong candidates before they cost you an interview slot.

The real on-call load. Not “participates in an on-call rotation”. Write: one week in four, roughly two pages per week, of which most are actionable. Senior SREs have been burned by vague rotations and read the omission as a warning. Being specific is a filter that works in your favour.

The honest current state. If you have no SLOs, no runbooks and alerting that fires on CPU rather than symptoms, say so. Candidates who want a greenfield reliability mandate find that attractive. Candidates who want to operate a mature platform self-select out, which is exactly what you want.

The authority the role carries. Can this person block a release? Declare an incident? Require a postmortem from another team? Write down the answer. If the answer is no to all three, revisit Step 1.

Step 3 — Source from adjacent pools, not from the title

Searching Dubai for the SRE title returns a pool of roughly 40 people, most of them employed and not looking. That was our first two weeks, and it produced three conversations.

The change that worked: we searched for backend and platform engineers who had owned production. Fourteen of the 23 credible candidates we eventually screened did not have SRE anywhere on their profile. The signals we searched for instead were on-call experience, incident command, capacity work, and Terraform or Kubernetes ownership in a production context rather than a lab one.

On geography, the regional pools that produced our strongest candidates were Egypt, Jordan, Pakistan and India, with a smaller stream of Europeans already in the Gulf. This tracks the broader market, where 80 to 85% of UAE tech hires come from international talent pools. Budget for relocation from the start rather than treating it as an exception.

Step 4 — Screen on incident reasoning, not tool inventory

This is the step that halved our interview time, and it is a single question asked well:

“Walk me through an outage you personally handled — from the moment you were paged to the moment the postmortem was closed.”

Then stay quiet and listen for four things.

Do they distinguish mitigation from resolution? Strong candidates restore service first and diagnose second, and they say so explicitly. Weak candidates describe a debugging session with users waiting.

Do they mention communication? Who did they tell, when, and what did they say while they still did not know the cause? Silence here is a serious signal — incident communication is most of the job in a small organisation.

Do they generalise the fix? “We added an alert for that specific condition” is a weak answer. “We realised we had no alerting on the class of failure, so we added symptom-based alerts on the user-facing path” is a strong one.

Do they take the blame or spread it? Blameless does not mean blame-free. Candidates who describe systemic causes without scapegoating individuals are describing a culture they can rebuild for you.

Twelve of our 23 candidates failed this conversation, including several with impeccable résumés and long tool lists. The tool list was never the constraint.

Step 5 — Run a 60-minute SLO and error-budget exercise

Give the candidate one of your real services — a short description, its dependencies, its traffic pattern — and ask them to propose service level objectives and defend them.

You are not grading the numbers. You are watching for whether they ask who the user is before proposing anything. An SRE who proposes 99.9% availability without asking whether this service is customer-facing or an internal batch job is pattern-matching, not reasoning.

The follow-up that separates seniors from mids: “We have burned 80% of the error budget with three weeks left in the quarter. The product team wants to ship a major release. What do you do?”

Weak answer: block the release. Also weak: approve it. Strong answer engages with the trade-off — what is in the release, does it touch the failing path, what is the cost of delay, what would we need to see to ship safely. That is the judgment you are actually buying.

Step 6 — Benchmark the offer against 2026 Dubai bands

Base salary, monthly, excluding allowances:

On top of base, budget housing allowance of AED 8,000 to 15,000 monthly, annual flight allowance, and end-of-service gratuity. With no personal income tax, net take-home runs 25 to 40% higher than an equivalent European package at the same nominal figure — worth stating explicitly to relocating candidates, who frequently mis-model it.

Add a dedicated on-call allowance of AED 2,000 to 5,000 monthly if the rotation is genuinely active. Candidates increasingly ask for this by name, and its absence reads as a signal that the rotation is unmanaged — which is often true.

Step 7 — Run visa processing in parallel and set a 90-day mandate

Start relocation paperwork at offer acceptance, not after the notice period ends. Those two clocks run in parallel and most employers waste three weeks running them in sequence.

Timelines: standard employment visa through MOHRE is 3 to 5 weeks including medical, Emirates ID and labour card. Golden Visa for specialised technical talent is 2 to 3 weeks with expedited processing. Free zone entities in DIFC, ADGM and Internet City have dedicated teams that shave a further 7 to 10 days. Total from open req to first day: 45 to 75 days.

Then write the 90-day mandate before they arrive. Ours was three items: SLOs defined and agreed for the top five user-facing services; alerting migrated from resource metrics to symptom-based; one incident run end to end with a published postmortem. All three are measurable, none require org-wide change, and together they answer the question that opened the role.

The 3 mistakes we made, so you can skip them

1. We searched for the title for two weeks. It produced three conversations from a pool of about 40 mostly-passive people. Adjacent-pool sourcing produced 118 contacts and 23 screens in the same amount of effort.

2. We ran a Kubernetes deep-dive as our technical round. It graded recall, not judgment, and it passed two candidates who later failed the incident conversation. We dropped it entirely and the quality of hires went up.

3. We under-specified incident authority in the offer. Our first SRE spent his second month discovering that “can declare an incident” was true in principle and contested in practice. Put it in writing, and tell the engineering leads before the person starts.

How this compares across the region

Reliability hiring plays out differently in each Gulf and Asian hub. Singapore's pool is deeper but repriced faster, which our colleagues at HireDeveloper.sg track in detail. Tokyo's constraint is demographic rather than competitive, and JapanDev covers what that does to on-call staffing.

See also our role pages for site reliability engineers, Kubernetes engineers and DevOps engineers in the UAE, and the August 19 Dubai Chambers & Nasscom agreement analysis for what is about to happen to senior salary bands.

Frequently Asked Questions