Senior Data Scientist (Live Probabilistic Modelling - Remote)

Greenfield models nobody has built yet, next to existing ones that need to get better.

Live probabilistic modelling where the quality of the output is the product itself.

A small but growing ML/DS team where one person owns an outcome end to end, for months at a time.

Our client builds real-time prediction systems that run while an outcome is still undecided. Their models produce continuously updating probabilities; how it ends, by what margin, and a spread of derived outcomes. Those numbers feed a live commercial product. When the models are right, it runs on its own. When they are not, a human steps in and corrects them by hand.

Reducing how often that has to happen is a fair description of this job.

What You’ll Be Doing

• Own projects end to end, usually one to three months each, from the research through to something running in production with your name on it
• Find where the current models fall apart in specific situations and work out why. This is actual research, not parameter sweeps
• Lift the quality of the predictions made before an event is live
• Build the setup-phase model from nothing. Before the main run begins, decisions get made that shift the picture
• Extend coverage as new areas come in - one is live, another is about to be, so ownership will grow
• Help with the data engineering your own work depends on, and build the monitoring that shows how the models behave in production
• Work directly with the people who review and override your models in real time, and with product and engineering. It is a startup, so you will talk to everyone

What You’ll Need

Essential

• 5+ years of industry experience, with a senior data scientist’s foundations: statistics, probability and machine learning you can reason from
• You have carried a project from research into production, and can talk honestly about the challenges you faced on the way
• Advanced Python: you regularly get a model out of a notebook and into production
• Patience with colleagues who have no statistics background and sometimes cannot articulate what they are asking for. You translate rather than correct
• You take requests from product and the people operating the system seriously, including the ones that cut against the model you would rather build

Bonus

• A background where getting the probabilities right genuinely mattered: forecasting, pricing, risk, insurance, demand or reliability modelling
• Real-time or streaming model experience, or MLOps exposure
• Time somewhere with mature machine learning practice that you can bring with you
• Experience of models whose output is consumed live by non-technical operators

The Person

You are here for the problem. You would rather be judged on whether the numbers hold up under live conditions than on how modern the stack looks. You can be the most statistically literate person in a conversation without making that everyone else’s problem. And you want genuine ownership: for a month or six, the project is yours, with help on hand.

An honest filter: If the last few years of your career have been almost entirely working on LLMs, this will not play to your strengths. The work is classical statistical modelling, with AI-assisted coding.

Remote-first, with offices in Europe. You need to be within roughly two to three hours of CET.