DS interview prep checklist for seniors

After years on the hiring side for senior data science roles, here is the prep checklist I wish more candidates followed. It is less about knowing more algorithms and more about showing judgement.

1. Metrics before models. For any product or case question, start by naming the decision the metric supports, the north-star metric, and two or three guardrails. Interviewers notice immediately when a candidate jumps to modelling before defining what success looks like, and they notice even more when the candidate can say what would make them stop a launch.

2. Experiments with teeth. Be ready to design an experiment end to end: unit of randomisation, why that unit, how you would handle network effects or spillover, what sample size you need and what you would do if the result is flat. The strongest answers I have seen name the failure modes before the interviewer asks about them.

3. System design for data science, not just software. Know where features are computed, how training data differs from serving data, how you would detect and respond to drift, and what you would monitor after launch. A design that ends at 'train the model' is unfinished.

4. Behavioural answers as decisions, not tasks. Use one concrete example per story and spend most of it on the decision you owned, the trade-off, and the outcome — not on describing the team or the project setup.

5. Practise out loud, timed. Senior candidates often fail not on knowledge but on pacing: they spend twenty minutes on the first question and rush the rest. A one-hour round rewards a complete, prioritised answer over a perfect but unfinished one.

What has helped you most in DS interviews — the technical rounds, or the product and behavioural ones?