GOV.UK Registers: - Probabilistic matching with active learning - Data modellingDigital identity: - Graph database of server logs for counter-fraud analysis - Recommender modelGOV.UK website - Pattern mining and process mining of clickstreams - Interpretable embeddings of clickstreams and web pages - Topic modelling with gradual change over time, of questions submitted to the Number 10 coronavirus press conferencesMiscellaneous: - Implementation of an optimal network algorithm for weighted maximum-mean subtrees - Supervision of projects by aspiring data scientists

Copy editing of mathematical and scientific papers in LaTeX, requiring excellent written English to minimise ambiguity, knowledge of relevant styles, attention to detail, and the ability to debug other people’s code.

Building the interim reporting data infrastructure for the smart metering rollout in SAS.• Working on a large, complex data set (smart metering) under government scrutiny.• Promoting best practices in data science within a delivery focused team such as closely coupling code with documentation, using test procedures, version control and stylistic conventions.• Identified, through rigorous testing, the cause of severe disruption to a data pipeline with direct impact on customers, and regulatory mandates, revealing significant potential savings in compensation and fines.• Developing the interim smart metering deployment data infrastructure and planning its integration with an enduring solution. Acting as the technical subject matter expert, working closely with a delivery manager. • Supporting a junior data scientist (graduate programme) with their skills development.• Communicated a data narrative concerning the importance of sequences to the interpretation of smart metering deployment data

• Brought a prototype Shiny app involving heterogeneous data to the point of deployment across the ministry, straightened out statistical glitches and liaised with service users to assess requirements and manage demand.• Catalogued a defunct survey dataset to expose its ~$3.5M value.• End-to-end R workflow, collaboration with a large team on GitHub.• Curated and documented a large, complex national survey database (tourism) linking ~60 time-series within a single reporting framework, for non-technical exploration.• Used an open-source software stack (R, Shiny, GitHub) to write packages for public release.• Peer-reviewed a colleague’s MTAGDP code (specialist programming for research combining and raking surveys to model GDP at a regional level).

• Pioneered the use of R for processing of larger-than-memory datasets, visualising police deployment patterns, integrating spatial demographic data into analytics, and web scraping.• Evaluated new and existing datasets for potential to inform policy development.• Produced software for a partner government organisation to increase the efficiency of scientific analysis and to meet accreditation requirements.• Provided training for end-users (scientists) through workshops and documentation.• Designed and built a data extraction layer to replaced dozens of redundant queries and spreadsheets (SAP BusinessObjects).• Curated and linked national speed camera and general road-user offence data in a reporting layer.• Liaised with the Automobile Association to provide extensive access to and understanding of police data to support their research programme.

• Designed a new database architecture to better facilitate time-series and transactional reporting.• Automated a data validation process, reducing three person-days per month to a couple of machine hours with machine accuracy.• Documented a data management system for re-development.• Assisted in modelling a pilot training pipeline with discrete event simulation software (Arena).• Curated military service records in order to manage recruitment and redundancyplanning.

• Developed a workflow model for testing scenarios and changes to business rules.• Integrated disparate datasets into a consistent reporting database, and future-proofed it against the integration of new work streams and expansion of reporting lines.• Optimised data processes that pushed the limits of the computing infrastructure, and that supported near-real-time refreshing.• Scoped and defined the high-level requirements of an upgrade to the data warehouse.• Curated real-time workflow data for managing teams that transcribed live court cases across the criminal justice system.• Conceptualised, created and maintained an information format suitable for high-frequency aggregation and reporting.