Shell Energy Retail Ltd (SERL) provided home energy to 1.3 million households and broadband to 0.5 million households in the UK. I led the data science function which consisted of a team of five Data Scientists. Sitting in the wider Data & Insights team, we integrated machine learning solutions within business processes to deliver a better customer experience by tailoring multiproduct offerings in home energy, broadband, smart home tech and renewable energy solutions. • Machine learning (Scikit-learn, Scipy.stats) – gained practical knowledge of the advantages and pitfalls of a wide range of techniques, as well as a strong grounding in the theoretical foundations; familiar with both supervised ML for predictive analytics using regression and classification (e.g. linear regression, logistic regression, random forests, SVM, XGBoost, ...) and unsupervised ML (k-means, other clustering, association rules, ...).• Model evaluation and validation – applied best practice and methods to evaluate and improve the performance of ML models (metrics, hyperparameter tuning, cross-validation); used explainability tools to interpret model outputs (SHAP, ...).• Assessing impact – conducted multiple A/B tests to evaluate the performance of prescriptive personalised marketing campaigns; retention campaigns reduced churn, giving substantial financial benefits.• ‘Rocket’ methodology – developed a framework for how machine learning projects were delivered across Shell Energy; this was a bespoke version of CRISP-DM, mapped across three stages of ‘Ignition’, ‘Countdown’ and ‘Lift-off’. The methodology was a key output from a high-profile initiative and was successful in underpinning the ways of working in the multiple cross-functional delivery teams.• Recruitment – built a fantastic team of Data Scientists from external hires, internal career changes, interns and contractors; in the team at Shell Energy average tenure was over five years, which is high for this sector.

• Python programming – applied object-oriented programming and software engineering principles to develop robust Python packages, including comprehensive unit testing and documentation, ensuring scalable and maintainable codebases.• Data analysis and preparation (Pandas, NumPy, SQL) – performed extensive data analysis and preparation, including cleaning, transforming and feature engineering, to ensure high-quality input for machine learning models.• Data visualisation (Seaborn, Matplotlib, Streamlit, Tableau) – created descriptive and diagnostic data visualisations to effectively communicate insights and support data-driven decision-making in a dynamic and interactive manner.• Agile working (Scrum, Kanban) – used agile methodologies and tools to chunk tasks and deliver incremental improvements; championed a continuous improvement mindset in sprint retrospectives to synthesise the best practice identified across the delivery teams.• Customer lifetime value modelling – built the behavioural module of a CLV model with the lead developer and the Finance team; this was part of the Exec's monthly KPI scorecard.• Coaching and mentoring – provided coaching and mentorship to team members, contributing to their professional growth and overall team success.• Working relationships – established good networks with stakeholders, leading to smooth delivery; facilitated constructive dialogue among team members to resolve conflicts.• Dissemination – presented to large audiences at internal seminars, external workshops and conferences to share key findings and raise the profile of the AI work stream.• Upskilling – delivered training sessions on both technical and non-technical subjects.• AI roadmap – maintained a funnel of project proposals working with senior leadership, considering ethical and legal factors; this used an innovative ‘Nine Box’ approach to capture key points for each idea, which were then assessed to prioritise based on feasibility and impact.

First Utility was a home energy and broadband provider, purchased by Shell in 2018. I built several predictive machine learning models and worked with stakeholders to deploy these across the business.• MLOps – utilised selected MLOps best practice for a pragmatic approach to versioning, testing, automation, reproducibility, deployment and monitoring across each of data, ML models and code.• Data science platform (Databricks, AWS, MLFlow, ...) – collaborated with experts on the setup of a shared platform, consisting of Databricks on AWS and other cloud tools; familiar with DevOps techniques to streamline workflows, including CI/CD pipelines, containerisation and orchestration.• Data science architecture – developed a conceptual architecture with seven layers for ML pipelines from source to consumption; this placed emphasis on reusable assets and common libraries in Python.• Manipulation of data pipelines (SQL, APIs, Airflow, ...) – worked in partnership with data warehouse teams on ETL to ingest data (Bronze) and then organise into intermediate processed assets (Silver) and serve to single versions of the truth (Gold); this included data processing at a large scale with assets comprising billions of data points.• Traditional NLP – supervised the delivery of sentiment analysis, text classification and text summarisation projects; the team built pipelines to extract data from 30,000 emails per month, resulting in improved efficiency in operational teams and substantial cost savings.• Forecasting (ARIMA, Prophet, …) – supervised the delivery of time series analysis projects for predicting the required resourcing in the customer service centre.

TRL is a leading organisation for independent transport research and consultancy.• Provided a robust evidence base to enable clients to make better informed policy decisions on the safety and effectiveness of sustainable travel interventions.• Helped clients to identify where transport investment should be targeted and prioritised to ensure the greatest financial benefit.• Delivered several high-profile projects as Technical Lead, instructing and working with effective multidisciplinary teams.• Wrote reports as the lead author and presented in various contexts, including at conferences.• Developed good working relationships with senior clients to clarify requirements and liaised closely with PMs to ensure delivery to time and budget, whilst taking decisions on technical matters.• Built regression models across a range of different problems, such as predicting the required level of cycle parking at stations, understanding the key predictors of car usage and estimating boarding times of different bus ticket types.• Undertook segmentation analysis of questionnaire data to predict the risk profile for individuals, based on demographics and attitudes; this involved factor analysis, cluster analysis and discriminant analysis.• Experience in a range of hypothesis testing techniques, such as T-Tests, Two Proportion Z-Tests and ANOVA.• Research interest on the reuse of the 'data exhaust' and 'digital footprint' from ubiquitous computing systems to gain new insights and also to understand how ICT can be used to encourage behaviour change.• Manipulated and analysed large datasets on rail passenger crowding.• Worked on a variety of projects in public transport, cycling and transport technology, gaining expertise in areas such as government policy, multimodal systems, passenger behaviour, smartphone apps, transport infrastructure, data integrity, system architecture, data markets, cost-benefit analysis, uptake demand modelling and impact assessments.

Modelling of various public transport initiatives as part of a wider transport assessment for a city council.

Data analysis of public transport surveys, typically before and after a change to assess its impact on journey times and sources of delay.