Turn complex data into decisions that matter.
Agile Bridge is looking for a Data Scientist who can move confidently from messy data and ambiguous business questions to reliable models, clear insight and measurable value. This is not a reporting-only role: you will build, evaluate and explain statistical and machine learning solutions, then work with engineering teams to prepare successful models for operational use.
What to expect in this role
You will work on business problems that require more than a dashboard or a standard software solution. Depending on the project, you may begin by investigating whether the available data can support a meaningful prediction, develop and compare several modelling approaches, or evaluate whether an existing model remains reliable. You will be expected to understand the decision the model must support, not only optimise a technical metric.
The role covers the data science lifecycle from problem framing and exploratory analysis through model development, validation and operational handover. You will retain ownership of the model's scientific validity and interpretation, while working closely with Data Engineers, AI Engineers and Platform teams on data pipelines, integration and production deployment.
What you will do
- Frame business problems as testable analytical or predictive questions with clear success criteria.
- Prepare and explore structured and unstructured data, resolve material quality issues and engineer useful features.
- Develop statistical and machine learning models using fit-for-purpose supervised or unsupervised techniques.
- Design rigorous validation approaches and evaluate performance, errors, robustness, bias and explainability.
- Translate findings into clear recommendations, visualisations and decision-focused narratives for technical and non-technical stakeholders.
- Collaborate with Data Engineering, AI Engineering and Platform teams to prepare validated models for production use.
- Monitor model performance and recommend recalibration, retraining, replacement or retirement when the evidence requires it.
- Explore LLM and RAG approaches where they genuinely improve analytical workflows or decision support.
What you need
- A bachelor's degree in Data Science, Statistics, Computer Science, Mathematics, Engineering or another relevant quantitative field.
- Typically 3-5 years' relevant experience developing and evaluating statistical or machine learning solutions.
- Strong Python and SQL capability.
- Sound knowledge of statistics, experiment design, feature engineering, model evaluation and data visualisation.
- Evidence that you can write reproducible code and communicate analytical trade-offs clearly.
- The judgement to choose an appropriate method, including knowing when a simpler solution is the better solution.
Core technical skills
We do not expect one person to have used every library or platform listed below. We do, however, expect strong capability in the core areas and evidence that you can select tools based on the problem rather than familiarity alone.
- Programming and querying: Strong Python and SQL skills, including the ability to write readable, testable and reusable analytical code.
- Data preparation: Hands-on experience with data manipulation and feature engineering using tools such as pandas, Polars or equivalent libraries.
- Machine learning: Practical experience developing supervised and unsupervised models using scikit-learn or comparable frameworks. This should include regression, classification and clustering; time-series or deep-learning exposure is advantageous where relevant.
- Statistics and experimentation: A sound understanding of probability, statistical inference, hypothesis testing, sampling, experimental design and the assumptions behind the methods you use.
- Model evaluation: The ability to select meaningful metrics, establish baselines, perform cross-validation and error analysis, test robustness and explain trade-offs between competing models.
- Visualisation and communication: Experience presenting findings using Matplotlib, Plotly, Power BI, Tableau or equivalent tools, with emphasis on clear interpretation rather than decorative reporting.
- Reproducible development: Working knowledge of Git, code review, notebooks and structured project practices that allow another person to reproduce, review and maintain your work.
Useful, but not essential
- An Honours or Master's degree in a relevant quantitative field.
- Exposure to a cloud data or machine learning platform such as Azure Machine Learning, AWS SageMaker, Google Vertex AI or Databricks.
- Experience with model tracking, packaging or monitoring tools such as MLflow, and an understanding of data drift, concept drift and retraining triggers.
- Experience working with distributed data platforms, ETL or ELT workflows, containers or production machine learning practices.
- Practical experience applying and evaluating generative AI, LLM or RAG solutions, including their limitations and responsible-use considerations.
How you work
- You are evidence-driven and comfortable challenging weak assumptions.
- You balance technical depth with commercial and operational reality.
- You explain uncertainty honestly and do not hide behind technical language.
- You collaborate well across business, data and engineering disciplines.
- You remain curious, but apply new technology selectively rather than for novelty.
The opportunity
You will join an innovation-focused environment where data science is expected to solve real problems, not produce experiments that never leave a notebook. The role offers scope to work across predictive analytics, responsible AI and emerging generative AI applications while partnering closely with experienced engineering teams.
How success will be measured
- Analytical work addresses a clearly defined business need and produces evidence that stakeholders can act on.
- Models are valid, explainable and appropriately governed for their intended use.
- Code, assumptions and experiments are reproducible and suitable for review or handover.
- Solutions continue to perform after implementation, with deterioration and emerging risks identified early.
