How GORS Careers Use AI in the Ministry of Justice

Operational research is one of those career names that can feel closed off until someone shows you what the work actually looks like. In a GOV.UK career story, Vlad does exactly that. His path through university, the Civil Service and the Ministry of Justice shows that this field is not only about numbers. It is about asking better questions, testing what works and making public services easier for people to use. For readers who are still working out where data, policy and public service meet, this matters. You do not have to begin with a perfect career plan. Sometimes you study one thing, enjoy one part of it more than expected, and only later find the job title that fits.

Vlad studied Security and Crime Science at University College London, a degree that brought together criminology, social research, psychology and coding. He also chose optional modules in web development and simulation, then spent six months on placement as a part-time Research Intern in the Mayor's Office for Policing and Crime. That mix is worth noticing. It shows how public sector data work often grows out of more than one subject, not a single narrow specialism. He says he did not hear about GORS until after joining the Civil Service. Once he did, the pieces started to line up. A lot of his degree had already touched operational research, especially the work of measuring outcomes after crime-prevention interventions. **What this means:** if you like evidence, social questions and practical problem-solving, you may already be closer to operational research than you think.

There is another part of the story that many students and early-career readers will recognise: mentorship matters. Vlad met a mentor in the Ministry of Justice through a charity while he was still at university. That mentor was a Project Manager rather than a GORS analyst, but the conversations still helped him understand what Civil Service work was actually like day to day. Sometimes the most useful guide is not someone with your exact job. It is someone who can make a workplace feel less mysterious. That helped shape a career built around real-world impact. Vlad says a large share of his degree focused on measuring what changes after an intervention. If you care about whether a policy actually helps people, rather than whether it only sounds good on paper, that mindset is a strong starting point for government analysis.

Today, Vlad works as an Associate Data Science Product Manager in Probation Data Science at the Ministry of Justice. His role is broader than writing code. He looks after product vision, ethics, stakeholder management, risk and documentation for data science projects, and he is currently focused on work involving large language models. In plain terms, he helps decide what a tool is for, how it should be used, what could go wrong and how to explain it clearly to the people who need to trust it. That mix of technical and people-facing work is a useful reminder for anyone thinking about AI careers. The job is not only to build a model. It is also to understand its limits, spot caveats early and translate complex ideas for non-technical colleagues. **What it means for you:** communication is not an optional extra in this kind of work. It sits right alongside analysis.

Before moving into product management, Vlad worked as a Senior Data Scientist in the same department. His projects included large-scale data linking across the criminal justice system, fines enforcement, reconviction rate estimation, extracting insight from employee data and prototyping a labour market dashboard. Read together, those examples tell us something important about Civil Service analytical roles: the work changes, the subject matter moves, and you keep building new skills as you go. Vlad's reflection on growth feels especially useful here. He explains that after a few projects, you look back and realise how much has changed, not just technically but in people skills, organisation and communication too. That is a healthy message for readers who feel they must be fully formed before they apply. Often, you become ready by doing the work, with support, practice and time.

One of the clearest examples from the GOV.UK case study is a tool called Contact Log Semantic Search. According to the Ministry of Justice story, probation staff write millions of reports each year about people on probation. These contact logs are unstructured, numerous and written in very different ways, which makes it hard for staff to find the right note when they need it for a risk assessment or a handover. Vlad explains that the team trained a large language model on large amounts of text so it could work out how words, phrases and sentences relate to each other. When a practitioner searches, the tool does not only hunt for an exact match. It reads the person's contact log and looks for similar meaning, then scores which entries are most relevant. **What this means:** if one person writes arson, another writes fire-setting and another uses a different but related phrase, the tool has a better chance of finding all the relevant records without making staff guess every possible search term.

Why does that matter? Because time spent wrestling with poor search tools is time not spent with people. Vlad says the goal is to return more relevant information, cut the time needed to search and reduce the need for repeated keyword attempts. In a probation setting, that can support better preparation, better risk assessment and more consistent case handovers. The public value is direct: if staff can find the right information sooner, they have more time for the work that supports rehabilitation and public safety. A second project shows a different problem that data science in government has to solve. Administrative data is messy. A person may not have one clear identifier across services, and they may appear under different names, addresses or aliases over time. That makes it harder to estimate how interventions are working or how someone interacts with other public services such as health, work and education. Vlad says he helped strengthen person-linking work by adding another data source to an existing Splink linkage. GOV.UK notes that Splink, an open-source package created by the Ministry of Justice data linking team, has been downloaded more than 10 million times worldwide.

If you are now wondering what it takes to get into GORS, Vlad's advice is reassuringly straightforward. The first formal requirement is that at least 50% of your degree should be numerate. Beyond that, he puts real weight on curiosity. Not on knowing the right programming language, and not on having a polished career story from the age of 18. He argues that analytical aptitude matters more: can you think logically, choose an appropriate method and explain your findings in a form other people can use? He also makes a practical point that many students need to hear. If internships do not come through, that does not mean you stop building evidence of your skills. Vlad suggests that time spent making your own coding projects with open data, or even with synthetic data, can teach you a great deal and strengthen your CV. We would add one final lesson from the whole case study: operational research is not hidden behind jargon. It is people using evidence, coding and clear thinking to help public services work better. Once you see that, the career path starts to feel much more open.