S.AI.F
Book a demo
Careers · 4 open positions

We sign our name to other people's models. Come hold the pen.

S.AI.F issues independent Trust Proofs for AI systems. That only works if the people running the tests are better than the people who built the model — and are free to say no. Independence is a hiring criterion, not a slogan.

See open positionsHow we hire
Abu Dhabi
HQ, with hybrid and on-premise client work
No build role
Assurance staff never build what they certify
Published work
Methods and findings go out under your name
Veto held
Auditors can block certification. Commercial cannot overrule it.
Open positions

Four roles, hiring now.

Select a role to read its full specification below, then apply with the form at the bottom of the page. Every posting states its level, its package and what it will be judged on.

Job specification · SAIF-ENG-01

Data & AI Model Engineer

Reports to
Head of Validation Engineering
Level & package
L3–L4 · competitive salary & benefits
Location
Abu Dhabi · hybrid
Clearance
Standard vetting + NDA

You build and run the machinery behind a Trust Proof: data quality gates, evaluation harnesses, fairness measurement and the evidence pipeline that makes a result re-runnable years later by someone who does not trust you. This is engineering in service of a signature — if a result cannot be reproduced from the artefacts alone, it does not count.

What you will own
Evaluation harnesses and frozen holdout sets, versioned so a result can be re-run years later
Data quality gates: completeness, duplication, label noise, train/test leakage, poisoning checks
Fairness and representation measurement across named groups, wired into the pipeline as a gate
The evidence pipeline — hashing artefacts on arrival and assembling what the certificate cites
Drift and fairness monitoring on live client traffic, with alerting that reaches a named owner
Essential
Strong Python and solid engineering discipline — tests, review, reproducible environments
Hands-on ML evaluation: holdout design, metric selection, groundedness and hallucination measurement
Data pipeline experience at scale, and the instinct that most model failures are data failures
Can explain a statistical result to a regulator without either lying or hiding behind jargon
Willing to be told your own pipeline is wrong, and to prove it either way
Desirable
Fairness tooling — Fairlearn, AIF360, Aequitas
Confidential compute or TEE attestation exposure
Arabic or multilingual evaluation experience
Published benchmarks, datasets or eval methodology
Judged on · first 12 months
PIPELINEEvery gate in your area runs unattended and is re-runnable from artefacts alone
THROUGHPUTAssessment time cut without loosening a single threshold
ESCAPESNo finding reaches a client report that the pipeline should have caught first
Tools you will use
PythonHELMlm-evaluation-harnessInspectGreat ExpectationsFairlearnMLflowOpenTelemetry
Engineering sits outside the assurance signing chain: you build the tests, auditors sign the result. You will never be asked to certify something you built.
Hiring process

Five steps, three weeks, one paid exercise.

Every candidate gets the same structured scorecard and written feedback, whether or not we make an offer. The technical exercise is paid at day rate because unpaid work is a tax on the people who can least afford it.

  1. STEP 01 · DAY 0
    Application

    Your CV through the form below, with a short message if you like. No cover letter.

  2. STEP 02 · DAY 3
    Screen

    30 minutes with the hiring lead on scope, band and independence.

  3. STEP 03 · WEEK 1
    Paid exercise

    A real, anonymised artefact. Four hours, paid, done on your own schedule.

  4. STEP 04 · WEEK 2
    Panel

    You defend your findings to two assessors and one non-technical reviewer.

  5. STEP 05 · WEEK 3
    Offer

    Package stated up front, no negotiation theatre. References after offer.

Conflict of interest. Assurance roles require you to declare current and recent employers, equity holdings and advisory positions in AI vendors. A conflict is not a rejection — an undeclared one is. You may be recused from individual engagements without effect on your role.
Salary & benefits

Competitive salary, and the conditions independence needs.

Benchmarked annually against Gulf market data for equivalent assurance and AI engineering roles, reviewed at the same time for everyone, and paid tax-free in the UAE. The package is stated at offer — we do not ask your current salary.

Pay
Competitive, benchmarked, tax-free
· Market-benchmarked base against Gulf assurance and AI engineering roles
· One review cycle for everyone, same criteria, no off-cycle deals
· Annual performance bonus tied to published objectives, never to revenue closed
· Founding equity for senior hires
Health & family
Covered, including dependants
· Private medical, dental and optical for you, your partner and children
· Life and income protection cover
· Parental leave beyond statutory minimum, equal for both parents
· Annual flights home and schooling support for relocated families
Time & place
Hybrid, with real leave
· 30 days annual leave plus UAE public holidays
· Hybrid by default; on-premise client work is planned, not sprung on you
· Relocation and visa sponsorship handled end to end
· No on-call rota for assurance staff — a rushed audit is a bad audit
Craft
Paid for, not squeezed in
· Certification fully funded — ISO 42001 lead auditor, CISA, security tracks
· Conference and research budget, with time to use it
· Publication encouraged under your own name, methods included
· Compute and lab access for red-team and evaluation research
Pay transparency. We publish the level with every posting, state the package at offer, and never ask what you earn now. Equivalent roles at the same level are paid within the same range — corrected at review if they drift apart.
Application

Apply in one pass.

Step 1 · Which role
Step 2 · About you
CV / résumé
Step 3 · In your own words
Two questions for this role. Both are optional, but a few honest lines here tell us more than any CV.
Step 4 · Declarations
Read the privacy notice.
Only your name, email and the two declarations are required. A CV helps us most.
Your application goes straight to the hiring team, and you receive a confirmation email with its reference.
Applying for
Data & AI Model Engineer
Builds the pipelines that test other people’s models — reproducibly, at scale, and without touching what it certifies.
Ref
SAIF-ENG-01
Openings
3 openings
Package
L3–L4 · competitive salary & benefits
Location
Abu Dhabi · hybrid
What we do with it
01Read by two people against a written scorecard.
02Answer within five business days, whichever way it goes.
03Written feedback if you reach the exercise stage.
04Held 12 months, then deleted. Withdraw any time.
Adjustments for disability or access needs: mention them in your message and we will adapt the process.