Define the role before the search
Start with the scientific or business problem and the maturity of your data. Specify whether you need exploratory research, model experimentation, data pipelines, inference infrastructure or ongoing model monitoring. Research scientist, data scientist and ML engineer are not interchangeable roles.
What to evaluate in candidates
Explore how the candidate framed a problem, established baselines, selected evaluation metrics and handled data quality or deployment constraints. For research roles, discuss reproducibility and contributions; for applied roles, discuss handoff, monitoring and maintenance.