Day-to-day operating brief
What Data Scientist work looks like in practice
A Data Scientist frames questions, explores data and builds statistical or machine-learning models to explain, predict or optimise outcomes. Unlike production-focused ML engineering, the role emphasises experiment design, inference, feature analysis and communication of uncertainty. Terms should cover data permission, notebooks and code, model IP, compute, peer review, publication, external collaboration and the boundary between exploratory findings and operational decisions. Research logs should preserve discarded hypotheses and parameter searches as well as the selected result. This helps reviewers distinguish planned evidence from chance findings discovered after repeated exploration of the same data. A recommendation should state the decision it informs and the cost of different errors. Identical headline accuracy can produce very different operational consequences when false positives and false negatives affect people unequally. Experiment review should record plausible alternative explanations and information unavailable at prediction time, preventing hindsight variables from entering a model intended for earlier decisions.
Individually written and reviewed against the maintained role record.