Data Scientist in London

Dyad is seeking a Data Scientist to help grow our analytical capabilities across our teams. This role fits someone who can pull, interrogate, and shape data from across the company, document the evaluations and benchmarks that matter to our AI Platform team and to Commercial, and turn all of it into dashboards, reports, and presentations that other people can act on. The role prioritises end of the data-science spectrum. It prioritises fluency with Python, SQL, and visualisation; clear reasoning about data quality and measurement; and communicating complex findings to stakeholders across the business. Communication fluency is a first-class requirement: a correct analysis that stakeholders cannot act on is a failure of the role, not of the audience. You will work across Commercial, AI Platform, and BetterLetter, reporting into the Chief Clinical Product Officer. This role is offered on a hybrid basis from our London office.

Core responsibilities

• Data extraction and analysis: Work with BetterLetter, AI Platform, QARA, and Commercial to pull data from production systems, customer environments, and internal tooling. Clean, join, aggregate, and interrogate datasets with rigour in order to communicate findings to all stakeholders. Flag where data is missing, unreliable, or not yet instrumented to support the question being asked, and recommend what to do about it.
• Dashboarding and reporting: Build and maintain dashboards for internal teams (product, commercial, leadership) and, where appropriate, customers. Produce recurring reports (customer-facing metrics, operational KPIs, board packs and investor updates as that becomes necessary) that are accurate, legible, and consistent over time. Run bespoke analyses to support sales, renewals, clinical conversations, and strategic decisions. Present findings clearly to non-technical audiences, including senior leadership and customers.
• Benchmarks and evaluations: Turn benchmark and evaluation outputs produced by the AI Platform team into documentation, reports, and visualisations that other teams can use. Communicate technical evaluation metrics in understandable ways, and describe how evaluation results change over time in terms non-specialists can act on.

Requirements

• Experience and background: A track record of applied data analysis work in a commercial setting is a must, with at least 3 years of experience; this is not a graduate role. We are seeking candidates with experience pulling, cleaning, and analysing data from production systems along with reporting and data visualisation. You should also be comfortable presenting findings to non-technical stakeholders, including senior leadership or customers. Experience working in or alongside teams building data-intensive products, ideally including ML or AI systems, is highly desirable. You might be trained as a data scientist with a preference for data work and strong applied data and statistical skills, or come from an analyst background but with sufficient fluency in writing Python to build and own reporting and analyses independently. Healthcare experience is a plus but not required.
• Technical skills: Python for data work: pandas, NumPy, Jupyter, plotting libraries (matplotlib, Plotly, seaborn), and enough general Python to write small tools and scripts without help. SQL across common dialects, including reading and reasoning about non-trivial queries and joins. A modern BI or dashboarding stack (Metabase, Looker, Superset, or equivalent), sufficient to build and maintain dashboards without engineering help for most work. Basic statistical thinking: sampling, confidence, effect sizes, and distinguishing a meaningful difference from noise. Reading and interpreting evaluation outputs from AI systems: precision and recall, error taxonomies, and what model metrics mean for a non-specialist audience.
• Personal attributes: Communication-led: treats clear presentation as part of the analysis, not an afterthought. Pragmatic and outcome-focused, willing to own the analytical question end-to-end. Comfortable flagging data-quality issues early and shaping the question rather than only answering it. Cross-functional by instinct: works effectively across engineering, AI, commercial, and clinical colleagues.