Data Scientist (409AE8E)
- Hiring from
- Poland
- Work type
- Hybrid
- Posted
- Oct 4, 2026
Is this job info correct?
Referment is working with a financial technology business that helps banks turn transaction data into useful customer insights and personalised digital banking experiences. Its platform combines transaction enrichment with financial intelligence, giving banks information they can use in customer relationship management, advisory services and AI agents.
As a Data Scientist, you'll work on the science behind that platform. Your focus may be transaction enrichment, customer intelligence or a combination of both: turning messy records into reliable merchant, category and location signals, then building behavioural features, scores and segments from those signals.
Banks use this work in production and need to explain it to auditors and regulators. You'll therefore build outputs that are documented, explainable and stable, working alongside senior data scientists who establish the standards and frameworks. Day to day, you'll use Python, SQL and dbt, an event warehouse such as ClickHouse, and Airflow pipelines, working with real banking data and synthetic datasets for quality assurance and demonstrations.
The Role
This is a B2B contract engagement with two office days and three remote days each week. The source lists equity participation, private healthcare, wellbeing and leave benefits, support for everyday working costs and team activities. It also describes international project exposure, career development and a focus on work-life balance.
This could suit a data scientist who has begun shipping production features or models and wants to deepen their work on transactional data, with senior scientific guidance and a clear emphasis on validation, reproducibility and explainability.
#Referment
As a Data Scientist, you'll work on the science behind that platform. Your focus may be transaction enrichment, customer intelligence or a combination of both: turning messy records into reliable merchant, category and location signals, then building behavioural features, scores and segments from those signals.
Banks use this work in production and need to explain it to auditors and regulators. You'll therefore build outputs that are documented, explainable and stable, working alongside senior data scientists who establish the standards and frameworks. Day to day, you'll use Python, SQL and dbt, an event warehouse such as ClickHouse, and Airflow pipelines, working with real banking data and synthetic datasets for quality assurance and demonstrations.
The Role
- Improve transaction classification and merchant-matching models. Contribute to quality, coverage and confidence metrics so banks can understand how much trust to place in each merchant, category and location signal.
- Turn banking event streams into reproducible financial features, including income stability, spending volatility, liquidity and balance trajectories. Work from specifications and frameworks set by senior colleagues, documenting features so they can be used reliably in models and rules.
- Contribute to financial-health, churn and propensity scores and behavioural segmentation. Support clear definitions, validation and drift checks, with explanations that risk and compliance stakeholders can understand.
- Implement and test statistical functions that banks use within their own rules and metrics. Examine behaviour on sparse or messy data and flag cases where calibration or explainability breaks down.
- Work with Product, CRM and Engineering to communicate what the analysis means and how it should be used, without relying on unnecessary jargon.
- At least two to three years' experience as a data scientist, with the ability to turn raw transaction, balance or event tables into a documented feature or validated score with guidance.
- Solid Python, including Pandas, and SQL skills. Your analysis should be reproducible by other people, rather than dependent on an undocumented personal workflow.
- Experience of an event or analytical warehouse such as ClickHouse or an equivalent, and practical familiarity with pipelines such as Airflow.
- Production machine learning experience, including model and drift monitoring. Relevant tools and approaches may include MLflow, feature stores and drift detection.
- Developing statistical judgement: you can explain why a method was chosen and respond to feedback when a rule or simple metric is more appropriate than a model.
- Some exposure to audit and explainability requirements, or a clear understanding of why definitions, lineage and avoiding silent data leakage matter.
- Full English proficiency and the ability to communicate clearly across technical and non-technical teams.
- Regular use of AI-assisted tools such as Cursor and Claude Code as part of your working practice.
- Banking, lending, cards, wealth or fintech, particularly work delivered on transaction data rather than only clickstream data.
- Hands-on delivery with dbt, Airflow or Spark beyond introductory familiarity.
- Real-time or event-driven scoring, fraud or credit-risk models, or financial-health modelling.
- Personalisation, next-best-action systems or marketing decisioning.
- Synthetic data for testing, including personas and edge cases. This is useful background rather than an entry requirement.
This is a B2B contract engagement with two office days and three remote days each week. The source lists equity participation, private healthcare, wellbeing and leave benefits, support for everyday working costs and team activities. It also describes international project exposure, career development and a focus on work-life balance.
This could suit a data scientist who has begun shipping production features or models and wants to deepen their work on transactional data, with senior scientific guidance and a clear emphasis on validation, reproducibility and explainability.
#Referment