Kpmg Data Scientist Interview Guide

We’ve seen KPMG evaluate data science candidates through an audit lens, not a pure modeling lens. Multiple candidates reported that the real test was whether they could connect technical work to business risk, especially when asked how machine learning could apply to auditing. That shows up again in the questions they ask: regression assumptions, multicollinearity, imbalanced data, and a spam classifier are all fairly standard, but the interviewers seem to care most about clear reasoning and practical relevance rather than pushing for research-level depth.

A recurring theme is that KPMG likes candidates who can explain their own projects crisply and defend the choices they made. One successful candidate described a final case presentation with both technical and non-technical interviewers where structure and communication mattered more than difficult coding. On the other hand, another candidate was surprised by how consulting-style the process became, with a substantial take-home centered on anomaly detection and clustering that felt much heavier than a junior role should. That contrast tells us the bar can shift depending on team and manager, but the common thread is business-facing data science.

We also see a company that can be slow and uneven in follow-up, so momentum is not always a signal of fit. The candidates who did well were the ones who could translate their background into audit use cases without overcomplicating things. In practice, KPMG seems to reward people who can stay grounded: explain the model, explain the tradeoff, and explain why it matters to the client.