Senior Analyst / Data Scientist, ProblemShared

• Fluent in SQL on a modern cloud warehouse. Strong on grain, joins, dedup, and small-sample reasoning
• Statistical literacy you can articulate — signal vs noise, significance, when a sample is too small to forecast from. Informed opinions on which models you would use and why
• A track record of visual data communications that were understood and acted upon without you in the room. Evidence informed decisions is a legacy you leave in your wake, not advice you have to re-explain
• Experience shipping EDA or reports or dashboards with at least two different visualisation tools, with a view on what work each is optimised for
• Comfortable working with stakeholders who don’t know what they need. The requirements emerge from discussion and feedback on your prototypes, not a static brief
• We know nobody ticks every box. If you have most of these, please apply

Desirable

• Join ProblemShared’s remote UK team delivering compassionate ADHD/Autism assessments. Flexible, supportive culture with development opportunities and real impact on patients
• Reporting to: Head of Analytics
• Our older reporting mechanisms work, but frequently through the manual toil of our office staff and some hard squinting over the results by our senior leaders
• We have built a data platform to replace all of this
• We are ingesting and synthesizing all of our organisations data, and we have begun rolling out automated dashboards and reports to close the last mile between the data and the decision
• This role sits at that pointy end, figuring out what the clinical and business stakeholders need to see so they can act, and then making sure they can see it
• Engage internal partners (clinical, ops, commercial, marketing, finance, HR) to surface the decisions behind the request, then build the visual data communications that support them. You can expect 2-3 new requests each week
• In the first instance you will answer these requests directly. Visualising and responding to the stakeholder through whatever tools get the insights to them fastest
• Where the decisions are recurring, you will work with our team to ship reliable and well governed dashboards or reports, democratising access to the insights so you don’t have to answer the question again every month. These are built in code, and peer reviewed
• You’ll push back when the question is the wrong question. Forecasting before the data supports it is hubris; small-sample inference is the bar. Tell us when not to build a dashboard
• You’ll work with our data engineers and software engineers to ensure that data quality issues or ommissions get resolved upstream, not in your dashboard. We’re building reproducible flows, not a house of cards
• What good looks like at 90 days
• Two recurring partner-grade reports run from the platform with no manual stitching, reviewed and signed off
• One of your peers has shipped visuals that you reviewed, and your expertise helped them improve it
• You’ve pushed back on at least one stakeholder request that was asking the wrong question
• Data quality issues you’re discovering are highlighted in your visuals (not concealed) and are flowing to our engineers via the data governance register
• Our 2026 priority is information legibility and decision support; the heavy statistics and ML problems on the horizon can wait their turn. You’ll collaborate with our other analysts and engineers, and mentor them where your skills allow. They will mentor you in return where their skills allow
• You’ll work with pseudonymised clinical and operational data. Information governance is part of how you design analyses, not a step bolted on at the end. Analyses live in version control. Dashboards are built in code. Nothing ships from a notebook straight to a stakeholder without peer review
• We use AI coding tools only when they’re additive and always with human oversight. As part of the central data team, you’ll help set the standards the rest of the company follows — so we expect you to know when to lean into LLMs and when to distrust them
• We hold a particular stance on how we represent the people in our data. Our patients are neurodivergent, and our framing of them is fit, not fault — dignity precedes identity, recognition over categorisation, the person is the primary recipient of anything we say about them. If those phrases land for you, we’ll probably get on
• Async is the default. Deep-work blocks are real and protected. Slack response cadence is hours, not minutes. Meeting load averages around two hours a day. Stakeholder feedback is an important part of your design process, but your deep work is what gets the next prototype shipped

• Short call (30 min) Two-way interview; we don’t want to waste your time. We’ll describe the role so you can determine your own interest. There will be some statistical reasoning and data visualisation questions
• Online Programming Test (15-60 minutes). Just SQL at the querying layer. No hard algorithms – we just need to know you find easy programming questions easy (without an AI)
• 2 Part Final Interview (60min/30min) – Technical conversation with the team, followed by a values conversation with the Chief Data Officer. Your choice whether these are back to back or on separate days – whatever is best for you