The second failed review is the one that changed how I screen for this. Both times the engineering was defensible. Both times what came back was about the journey: what happens when somebody abandons the sign-in halfway, what happens when they do not have the identity app on the phone they are holding, and what happens the second time a returning user signs in and the matching logic quietly creates a new account. We had hired for protocol knowledge and been beaten by product decisions nobody had made.
What This Role Actually Is
UAE PASS is the national digital identity used to sign in to government services and a growing set of private ones, and it also carries a document signing capability. Technically the authentication side is OpenID Connect layered on OAuth 2.0, which means the protocol itself holds no surprises for a senior engineer. The official reference is docs.uaepass.ae, and the service itself sits at uaepass.ae.
So if the protocol is ordinary, why hire against it specifically? Because access is gated rather than self-service. You do not create an account and start calling a production endpoint. The path runs through approval of your use case, then staging credentials, then an assessment of your integration covering both security and the user journey, and only then production. That sequence is the part that costs calendar time, and it is the part your screening should be aimed at. The mobile application requirements page is worth reading before you interview anybody, if only so you can tell when an answer is confident and wrong.
Our expert take
The pattern we see with any gated national integration, and this is not unique to the UAE, is that employers screen for the thing they can test easily and get surprised by the thing that gates them. Token validation is easy to test in an interview, so everybody tests it, and almost nobody fails a review on it. Account linking and abandoned journeys are awkward to test, so they get skipped, and that is where submissions come back. The whole value of a structured screen here is that it forces you to ask the uncomfortable questions in week one rather than month three.
Step 1: Decide Which Capability You Need, in One Sentence
Write it down before you write the job description. There are three materially different versions of this hire and the screening questions barely overlap.
Authentication only means you want users to sign in with their national identity and you will keep your own profile data. This is the cheapest version and the one most products actually need. Authentication plus profile means you intend to consume identity attributes and store them, which drags in data retention and accuracy questions you now own. Authentication plus signing means documents get signed through the service, which is a different capability with its own review considerations and a different class of legal expectation attached to the output.
The failure I see is a job description that says all three because nobody wanted to choose. You then interview for the union of the requirements, reject people who could have done the job, and pay a premium for the signing experience you will not use for eighteen months.
Step 2: Write the Role Against the Approval Gate
Put the sequence in the posting: use case approval, staging credentials, build against staging, assessment of the integration, production access. Name it as the thing the role owns.
This does two useful things. It filters, because engineers who have been through a gated onboarding recognise the shape of it immediately and the ones who have not tend to apply for something else. And it sets the expectation internally, which matters more than it sounds: a hiring manager who has written that sentence will not promise a stakeholder a four week delivery, because the sentence itself makes the dependency visible.
One phrasing that has worked for us: “you will own our integration from the first technical session through to production approval, including the evidence we submit for assessment”. Candidates who have done it ask what our use case is. Candidates who have not ask which library we use.
Step 3: Screen the Server Side Exchange, Not the Library
This is the five minute question that does the most work. Describe your architecture at a high level and ask two things: where does the authorization code get exchanged for tokens, and where does the identity token get validated.
The only acceptable answer puts both on your backend. If a candidate puts the code exchange in the browser, or validates the identity token in the mobile client and trusts the result, you have learned enough. This is not a trick question and it is not obscure, which is exactly why it is useful: it is the single most common place where an otherwise competent engineer reveals that their experience is with social sign-in buttons rather than with an identity an institution relies on.
Follow up with one probe: what do you check in the token beyond the signature. You are listening for the audience, the issuer, expiry and the nonce. A candidate who names three of those four has done this properly at least once.
Step 4: Treat the Mobile Flow as a Separate Skill
If your product has a mobile app, the app to app flow is a second screen rather than an extension of the first. The failure modes do not overlap with the web flow at all.
Two concrete things to ask. First, the identity app has to hand control back to your application, which means your application needs its own registered URI scheme, and using the sample scheme from the documentation in either staging or production is a known way to fail. A candidate who has shipped this will mention the scheme unprompted. Second, and this is the one that catches people: what does your app do when the user does not have the identity app installed. There is a designed answer involving a fallback and a clear instruction to the user, and there is an undesigned answer where the user taps a button and nothing happens. The review will find the second one.
Step 5: Test the Identity Edge Cases
These four questions are the heart of the method. None of them are about cryptography and all of them have failed a real submission.
- Assurance level. Not every account carries the same level of verification. Ask what the application does when the returned profile does not meet the level your service requires. The answer you want describes a specific user-facing path, not an error page.
- Residents, citizens and visitors. Your user base is not uniform. Ask how the integration behaves for an account type your service was not designed around, and whether that is a hard block or a degraded journey.
- Names. Arabic transliteration, name order and multi-part names will not match what is already in your database. Ask how they would compare an incoming identity to an existing record. If the answer is a string comparison on full name, you have found your duplicate account bug in advance.
- Linking. This is the one that cost us a review. A user who already has an email and password account in your system signs in with national identity for the first time. What happens? There is a right answer involving a deliberate, verified link and a wrong answer involving a second account with half the user’s history.
A candidate who handles three of these four has done the work. A candidate who handles all four and volunteers a fifth has probably failed a review themselves, which in my experience is the strongest signal on offer.
Want these four questions asked for you
Send us your use case and the account types you serve, and we will screen candidates against the edge cases your integration will be assessed on.
Get startedStep 6: Give a Staging Shaped Take Home
You cannot hand a candidate production credentials, and you should not hand them your staging credentials either. That constraint is an opportunity rather than a problem.
Set the task as a design document plus a working mock. Ask for a sequence diagram of the sign-in flow naming where each secret lives, a written list of every assumption they had to make because they could not see the real environment, and a small running implementation against a stub they write themselves. Cap it at three hours and say so.
What you are grading is the assumptions list. The job is largely about working against an environment you only partly control and documenting what you believe to be true so a reviewer can check it. A candidate whose assumptions list is specific and hedged in the right places will handle the assessment. A candidate who submits a working mock with no assumptions listed has either not noticed the gaps or has decided not to mention them, and both are expensive later. We use the same structure when hiring for other gated integrations, including the one described in our method for an e-invoicing integration developer in the UAE.
Step 7: Audit the First Go-Live Against the Screen
Once the integration is through, sit down for an hour with whatever came back during the assessment and compare it against what you asked in the interview. Three columns: what the review raised, whether the screen tested for it, and whether the person you hired handled it well.
The useful finding is almost never that you hired badly. It is that one of your questions was theatre and one of the things that mattered was never asked. Rewrite the screen, and keep the rewritten version with the job description rather than in somebody’s head. Two cycles of this produces a filter specific to your product rather than a generic identity interview, and that is the point at which hiring for this stops feeling like luck.
| Step | Time cost | The failure it prevents |
|---|---|---|
| 1. One sentence on capability | 20 minutes, once | Paying a premium for signing you will not ship |
| 2. Role written against the gate | 30 minutes | A delivery date promised without the approval path |
| 3. Server side exchange question | 5 minutes per candidate | Social sign-in experience mistaken for identity work |
| 4. Separate mobile screen | 10 minutes per candidate | A button that does nothing without the identity app |
| 5. Four edge case questions | 20 minutes per candidate | Duplicate accounts and undesigned journeys |
| 6. Staging shaped take home | 3 hours theirs, 40 min yours | Unstated assumptions discovered during assessment |
| 7. Post go-live audit | One hour, once per launch | A screen that never improves |
Our expert take
The question we get asked most often is whether to hire for this at all or to add it to an existing engineer’s plate. Our honest answer depends on one thing: whether the launch date was set by you or by somebody else. An internal date can absorb a submission coming back for changes. A date tied to a tender, a government service commitment or a partner launch cannot, and that is the only situation where we would argue for paying the premium. The code is not the scarce thing. Passing the review on the first attempt is.
When a Strong Generalist Is the Better Hire
I would rather lose this placement than have you overbuy, so here is the case against the specialist.
If your integration is authentication only, your product is not public sector facing, and your launch date has four weeks of slack in it, hire a senior backend engineer who knows OpenID Connect properly and give them the documentation. The protocol work is genuinely ordinary. Budget two extra weeks for the approval path and accept that your first submission may come back once. You will spend less overall, and the engineer you hire will be useful for the other eleven months of the year.
The specialist earns their premium in exactly three situations: a fixed external launch date, a signing requirement, or an existing user base you have to link to without creating duplicates. If none of those apply, step one has already answered your question. Where it does apply, the sequencing of the wider build matters as much as the hire, and our guide to building a government services app in the UAE covers the parts that sit either side of the integration.
The 3 Things This Will Not Fix
- An unapproved use case. No hire accelerates a decision that sits with somebody else. If your use case has not been accepted, the engineer you hire will spend their first month waiting, and the honest move is to delay the start date rather than to fill the time with adjacent work.
- A product that has not decided who its users are. Half the edge cases in step five are product questions. If nobody can tell you whether visitors are in scope, your engineer will make that decision by accident, in code, and you will discover it during assessment.
- Weak handling of regulated data generally. An identity integration raises the stakes on everything downstream. If your access control and audit trail are thin, this hire makes the problem visible rather than smaller. The screening approach we use for that class of problem is set out in our method for a PCI DSS payment engineer in Dubai, and the logic transfers.
Frequently Asked Questions
Is a UAE PASS integration a specialist hire or a normal backend task?
Can I hire this role remotely or does it need somebody in the UAE?
What does this role cost compared to a normal senior backend engineer?
What is the single most common reason an integration fails its review?
Screen for the review, not the framework
We shortlist UAE integration engineers against the edge cases that decide a go-live, and we will tell you when a generalist is the better call.
Get started