Senior Data Scientist

The Government Digital Service (GDS) is the digital centre of government - we are responsible for setting, leading and delivering the vision for a modern digital government.

Our priorities are to drive a modern digital government, by:

• joining up public sector services
• harnessing the power of AI for the public good
• strengthening and extending our digital and data public infrastructure
• elevating leadership and investing in talent
• funding for outcomes and procuring for growth and innovation
• committing to transparency and driving accountability

We are home to the Incubator for Artificial Intelligence (I.AI), the world-leading GOV.UK and at the forefront of coordinating the UK’s geospatial strategy and activity. We lead the Government Digital and Data function and champion the work of digital teams across government.

We’re part of the Department for Science, Innovation and Technology (DSIT) and employ more than 1,000 people all over the UK, with hubs in Manchester, London and Bristol.

The Government Digital Service is where talent translates into impact. From your first day, you’ll be working with some of the world’s most highly‑skilled digital professionals, all contributing their knowledge to make change on a national scale.

Join us for rewarding work that makes a difference across the UK. You'll solve some of the nation’s highest-priority digital challenges, helping millions of people access services they need

Former DSIT recruitment campaigns are continuing as usual, but candidates should be aware that following the Government’s announcement on the changes to some civil service departments, roles will be subject to the machinery of government moves and will ultimately be in one of the new departments. We will provide more information if you are selected for a role. This work remains of high importance to the civil service, and we thank you for your continued interest.

GDS is helping government get real, responsible value from AI. As a Senior Data Scientist in the GDS Product and Services’ AI programme, you'll turn fast-moving AI capability into tested, trustworthy services that improve how people experience government.

You'll work with product, delivery, user research, design and engineering colleagues to find where AI can help, check whether the underlying data and content are good enough, and design and run pilots that prove through rigorous testing what works, what doesn’t and why. That includes building evaluation methods for accuracy, safety and user value, and testing agentic approaches in controlled pilots before they reach the public. Your findings will shape how GDS and the rest of government adopt AI.

About the team

The AI programme sits within the GDS Products and Services directorate. It supports government to adopt AI safely, effectively and at scale. The team builds and improves GOV.UK Chat, which has been through large‑scale public pilots, red‑team and security testing, and a launch in the GOV.UK app. It is also experimenting with agentic AI, including how an assistant could take simple actions on a person’s behalf. The team trials new tools, builds shared capability and guidance, and works with departments on practical use cases that meet public sector standards for security, transparency and ethics.

You can read about some of our work here:

• https://gds.blog.gov.uk/2026/05/14/gov-uk-chat-launches/
• https://insidegovuk.blog.gov.uk/2026/05/15/developing-gov-uk-chat-our-data-science-and-ai-engineering-journey/

As a Senior Data Scientist, you’ll:

• lead technical work on evaluation and experimentation within the AI programme, working with product, delivery, user research, design and engineering colleagues to build AI products that meet user needs
• prototype and assess new and emerging AI tools, including agent-based systems, and make evidence-based recommendations on whether they’re worth adopting for government use cases
• shape problems with product and technical leads, helping to define scope and success measures, then design and run experiments and pilots using generative AI, agentic approaches and automation
• design robust, repeatable evaluation frameworks that show how well a tool or solution performs, where it fails and what risks it carries, including red‑team and guardrails
• judge whether datasets and content are suitable and representative to use for a given purpose, and advise on gaps and remedies
• write and review secure, maintainable, well‑tested code, and promote good engineering practice in data science work
• explain results, trade‑offs and recommendations to audiences from engineers to senior officials, so decisions rest on evidence
• apply government standards on data protection, security, transparency and AI ethics, and contribute to responsible AI practice in the team and the wider Data and AI community
• mentor and coach data scientists and share good practice, helping to raise capability in the team

Person specification

We’re interested in people who:

• a track record of applying data science and generative AI methods to real operational or user problems, working in multidisciplinary teams
• sound analytical rigour: you've evaluated models, tools or experiments, including LLM or RAG systems, using methods such as benchmark datasets, human or LLM‑based grading and red‑teaming, and reached clear, defensible conclusions that shape product delivery
• the ability to explain technical findings, limitations and uncertainty to non‑technical decision‑makers
• good judgement on responsible use of data and AI, including governance, privacy, security and ethics
• strong engineering habits: Python, clean and tested code, CI/CD, version control, and practical use of AI‑assisted coding tools to improve quality and speed
• hands‑on experience building LLM‑powered applications and AI agents, including retrieval‑augmented generation (RAG), with an understanding of embeddings, vector search and retrieval quality, using modern AI development frameworks (e.g., Amazon Strands, LangChain, LangGraph), and cloud AI platforms (preferably AWS Bedrock)
• understanding of agentic design patterns, such as tool and function calling, memory and multi‑agent set‑ups, and of how to test them safely.
• experience in the public sector, regulated environments, research or fast‑moving start‑ups.
• context engineering experience across commercial and open‑source models.