Build the function. Be its first practitioner.
Zero models in production. That is the honest starting line, and it is also the offer.
DigitalZone has the parts most companies are still fighting for: a unified warehouse, data engineering and analytics teams already running, and behavioural and transaction data from a digital goods business, a ticketing platform and a regulated trading venue. What it does not have is a single model deciding anything. Nothing scores fraud risk, nothing forecasts demand, nothing chooses what a customer sees next.
You are the first data science hire. You will build the team, set how it works, and personally be the data scientist for Platform: payments, identity and marketing technology. You report directly to the executive who owns Product and Data, so the distance between your finding and a decision is one conversation. It also means there is nobody to hide behind when a model does not work.
The hard part
Fraud in a cash-heavy market with agent networks and chargeback exposure does not sit still. Your model will be adversarial from day one, and the threshold you set is a live tradeoff between loss prevented and customers blocked. That call will be yours to make and yours to defend.
Identity is the other half. Customer records arrive from several products and a partner login, and resolving them into one trustworthy profile is what makes personalisation, risk scoring and half the company's roadmap possible. Nobody has done it yet.
What you'll do
- Build the function from one person to a team: charter, hiring, and the standard for how a model is specified, validated, shipped, monitored and retired.
- Own fraud detection as a product: transaction and payment risk scoring, chargeback and refund abuse, account takeover, promo abuse, agent and merchant anomaly detection. Set the thresholds with fraud and operations, and defend the loss-versus-friction tradeoff.
- Own the modelling side of identity management: identity resolution and record linkage across sources, deduplication and match confidence, profile completeness scoring, and the unified customer profile that both personalisation and risk read from.
- Drive the next-best-action and segmentation layer behind personalised customer communication.
- Get models into production and keep them there, with the cloud team on serving and MLOps: monitoring, drift, retraining, rollback, and one measurable business outcome per model you are willing to be judged on.
- Own the planning models: demand and stock forecasting, predictive budget, category and SKU level targets, churn and lifetime value.
- Own experiment design across product and marketing: hypothesis, minimum detectable effect, decision rule, and honest readouts including the null ones.
- Agree the boundary with data engineering and analytics. It is not settled today. Settling it is part of the job.
- Bring stakeholders a decision, not a notebook. Definitions come from the company metric contract, and an unvalidated figure does not become a fact by appearing in a deck.
Requirements
What you'll bring
- 7+ years in data science, 2+ leading or mentoring scientists. Having been the first or second scientist somewhere counts double.
- Models in production with a measured business outcome. Hard requirement: live traffic, monitored, effect quantified. A portfolio of notebooks is not this.
- Fraud or risk modelling depth, plus at least one of: identity resolution and entity matching, forecasting, churn and lifetime value, personalisation, uplift modelling, causal inference.
- Strong SQL against large transaction and event data in a columnar warehouse, strong Python, and code someone else can run next quarter.
- Experimental and causal rigour, including quasi-experimental methods where randomisation is not possible.
- Consumer fintech or payments fluency: transactions, funnels, cohorts, fraud patterns.
- The ability to align product, engineering, marketing and commercial on one problem definition before any modelling starts.
- AI-assisted working, with the rule that you never ship output you cannot explain.
Nice to have
- Built a data science function from zero, including the first hires and the first deployment path.
- Fraud or payment risk in a market with cash-heavy behaviour, chargeback exposure or agent networks.
- Identity graph, KYC or customer data platform experience.
- MENA or emerging-market consumer data, with the data quality reality that comes with it.
- Arabic.
What we will not pretend
There is no team yet, no serving infrastructure you can lean on, and no data science rubric to hide behind. The first six months are you, a warehouse and a lot of stakeholder conversations. What you get in return is a blank page at a company that has already decided models matter, and the rare chance to be judged on outcomes you defined yourself.
If you want to join a mature data science org, this is the wrong role. If you want to be the reason one exists, apply.
Benefits
- Immediate, large-scale impact on a high-growth business
- Top-of-the-market compensation packages
- Work alongside top regional talent, with team members from Talabat, Careem, Etisalat, and more