Senior Quantitative Researcher/Scientist

Oxford Data Plan•Oxford, Asia

About The Position

We are looking for a Senior Quantitative Researcher to take a leading role in developing and governing the statistical methodology behind our daily KPI indices. Our methodology determines how alternative data is transformed into robust estimates of company performance, including how models are specified, validated and governed before publication. It is developed and maintained by the OXDS team and runs every day across several hundred published indices. This is a new, dedicated senior position created to advance that methodology, introducing deeper statistical modelling, stronger uncertainty quantification, more robust approaches for sparse and heterogeneous data, and methods that continue to perform as our data sources and products evolve. In this role, you will be the senior statistical voice for the index production system, working within the OXDS team alongside software engineers and with data scientists across the product teams.

Requirements

5+ years of work experience in applied statistics, econometrics or quantitative data science or a PhD in a relevant field.
Demonstrated experience owning statistical methodology in a production system where modelling decisions reached external users.
Postgraduate training in statistics, econometrics or another quantitative discipline, or equivalent applied experience.
Deep understanding of regression modelling and statistical inference, including diagnostics, constrained estimation and uncertainty estimation.
Strong time-series expertise, including forecasting, seasonality, calendar effects, structural breaks and time-series cross-validation.
Strong understanding and experience with two or more of the following: Bayesian modelling, particularly hierarchical models; Kalman filtering; Errors-in-variables models; Change point detection; Conformal prediction; Forecast combination (e.g. Bates and Granger method).
Experience with sampling and panel methodology, including selection bias, unbalanced panels, reweighting and representativeness.
Experience calibrating thresholds and uncertainty measures for automated decision systems, with sound judgement around false positives and missed errors.
Ability to reason about methodology under production constraints such as data revisions, missing inputs, short histories and backward compatibility of published series.
Fluent Python with the statistical stack, including statsmodels, scipy, pandas and NumPy.
Comfortable contributing to a large shared production codebase using version control, code review and automated testing.
Experience writing statistical code that runs unattended, with appropriate validation, reproducibility and failure handling.
Strong SQL skills.
Excellent written communication and the ability to explain statistical reasoning clearly to non-statistical audiences.
Comfortable acting as the methodological authority within a multidisciplinary team and challenging weak approaches with clear reasoning.
Fluent in English.

Nice To Haves

Experience with alternative data, nowcasting or KPI/revenue estimation using sources such as transaction, web or app data.
Experience with uncertainty quantification for time series, including prediction intervals, conformal or adaptive conformal methods.
Experience with one or more advanced modelling approaches, such as Bayesian modelling, state-space models, hierarchical or pooled modelling, forecast reconciliation, or other relevant statistical methods.
Experience with index construction methodology, including chaining and rebasing.

Responsibilities

Lead the statistical methodology of index production end to end: model specification, estimation, adjustment, validation and forecasting.
Propose and develop substantial methodological enhancements to improve the quality of our KPI models beyond our existing approaches.
Own the adjustment and drift-control methodology: scaling, bounds, clipping and the treatment of outstanding corrections, including how thresholds and windows are selected and evidenced.
Own the statistical treatment of panel and source data before modelling, including coverage, reporting lag, sample drift and representativeness.
Set standards for the model lifecycle, including specification, validation, backtesting, retraining, and evaluation using only the information available at the time of publication.
Own the revision methodology for published series, including when restatement is justified and what historical consistency it should preserve.
Own the statistical logic behind production quality assurance, including input monitoring, outlier detection, alert calibration, and distinguishing data issues from model issues.
Maintain clear methodology specifications that engineers and data scientists can build against, and review methodological changes proposed by product teams.
Validate numerical equivalence when statistical libraries or scientific dependencies are upgraded.
Implement methodology as tested, production-quality Python and take accountability for the statistical correctness of published outputs.
Deliver training sessions to other data scientists to upskill them in our methodology, especially for new methods you have developed.
Mentor data scientists and raise statistical standards across the wider team.