Senior Data Scientist — Clean Cooking (PAYG LPG)

Role Overview

This is a senior individual contributor position inside our pay-as-you-go LPG business, where the raw material is a stream of commercial questions: which customers are about to stop refilling, what a price change will do to demand, which households are worth acquiring, and where the cost-to-serve is quietly leaking. You will turn those questions into models and data products that sales, operations and commercial leaders actually use to make decisions — not artefacts that stay in a notebook. The job blends fieldwork with technical depth: part of your time is spent with agents and customers in our markets, and the rest is spent framing statistical problems, building and validating models, and handing them to production in partnership with our engineering teams.

Your performance is judged on the commercial metrics of the business, not on model scores in isolation. The role is deliberately scoped as a hands-on senior IC track rather than a management post, and it is not an MLOps or BI position — pipeline and deployment infrastructure belongs to dedicated engineering, and dashboarding sits with the analytics team. What you own is the thinking between a business problem and a deployed, monitored model.

Key Responsibilities

• Build first-hand knowledge of customer and commercial realities across our markets, including time in the field alongside sales agents and the customers they serve.
• Convert loosely stated business problems into well-framed statistical questions, then present findings credibly to audiences ranging from fellow practitioners to BU leadership.
• Prioritise your own portfolio aggressively — deciding where a data product will genuinely shift activation, refill frequency, retention or unit economics, and saying plainly where it will not.
• Design, build and evaluate machine learning models for churn and dormancy prediction, credit and payment-behaviour modelling, demand forecasting, anomaly detection and customer segmentation.
• Apply probabilistic and Bayesian techniques to quantify uncertainty around pricing elasticity, promotion uplift and marketing/media effectiveness, so decisions can be made with the risk made explicit.
• Design and analyse experiments — A/B, geo and quasi-experimental designs — under field conditions where clean randomisation is often unavailable.
• Carry out rigorous exploratory analysis, feature engineering and data wrangling across large structured and semi-structured datasets, keeping code reproducible, documented and consistent with team engineering standards.
• Own each model from framing through a validated, deployment-ready handoff with data and analytics engineering, then jointly own its live monitoring — tracking performance, spotting degradation and recommending retraining or redesign when triggers are hit.

Requirements & Qualifications

• A degree in Computer Science, Statistics, Mathematics, Engineering, Economics or another strongly quantitative discipline. An advanced degree is welcome but not required; demonstrable shipped impact carries more weight.
• Five to eight years of hands-on data science or applied ML experience, including at least two years owning data products end to end.
• A track record of data products that measurably moved a business outcome — you can walk through the problem, the model, the decision it changed and the size of the effect.
• Strong command of classical ML — gradient boosting, regression, clustering, ranking and time-series forecasting — together with the judgment to recognise when a simple approach outperforms a sophisticated one.
• Solid grounding in probabilistic modelling, Bayesian inference and uncertainty quantification, with practical experience in a probabilistic programming language such as PyMC or Stan.
• High proficiency in Python and its standard scientific stack, plus strong SQL covering complex multi-table queries and window functions.
• Deep familiarity with model evaluation: cross-validation, calibration and the discipline to choose business-aligned metrics over convenient ones.
• Experience designing experiments and applying statistical hypothesis testing.
• Comfort working with cloud data warehouses and experiment-tracking tooling (we use AWS and MLflow; equivalent tools are perfectly acceptable).
• Strong communication and storytelling — the ability to hold the attention of a room of non-technical commercial leaders — alongside real empathy for customers and colleagues, and the habit of estimating a proposal's P&L impact before building it.
• Confidence managing upwards: surfacing product needs, trade-offs and timelines rather than absorbing them silently.
• Strongly preferred: experience in PAYG, fintech lending, telco or comparable emerging-market consumer businesses, including an understanding of irregular incomes, mobile-money payment behaviour and thin, messy data.
• Nice to have: survival modelling, causal inference or marketing mix modelling; operations research or optimisation exposure relevant to last-mile routing and scheduling; familiarity with MLOps and model deployment on AWS (SageMaker, Lambda, ECS).

What We Offer / Why Join

• A senior individual contributor mandate with real ownership: you hold the thread from problem framing to production handoff, rather than being a downstream consumer of someone else's infrastructure.
• Direct exposure to influence — your analysis and recommendations reach BU leadership and shape pricing, promotion and retention decisions.
• Genuine field contact with agents and customers, so your models are grounded in how the business actually operates.
• Clear engineering partnerships: data and analytics engineering own pipelines, and MLOps owns deployment infrastructure, leaving you free to focus on modelling and measurement.
• Room to grow — as the data science function expands, a path to leading a small team may open.