Lead Data Scientist

About the role

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.

Responsibilities

• - 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.

Skills

• SQL
• Python
• MLOps
• fraud modelling
• risk modelling
• identity resolution
• entity matching
• forecasting
• churn and lifetime value modelling
• personalisation
• causal inference
• experimental design
• transaction risk analysis
• production model deployment