Begin with the platform problem
Describe what must become more reliable, accessible or scalable in the next 12 months. A greenfield platform, a migration and an established production estate each require different evidence.
Be explicit about where the role sits between software engineering, analytics engineering, architecture and operations. Ambiguity here is one of the fastest ways to attract a large but irrelevant field.
Separate essentials from preferences
Choose the few capabilities that are genuinely required on day one. Cloud vendor exposure can transfer; experience making sound decisions about modelling, orchestration, observability and cost is often more durable.
State what the person can learn after joining. This makes the proposition more credible and avoids excluding strong engineers whose exact toolset differs.
Assess decisions, not vocabulary
Ask candidates to explain a comparable system: the constraints, alternatives considered, failure modes and trade-offs. Good assessment reveals how someone reasons when reliability, speed, governance and cost pull in different directions.
Use a consistent scorecard covering engineering depth, delivery, collaboration and the outcomes required in your environment.
Run a process that respects specialists
Explain the mandate and interview stages early. Keep technical exercises proportionate, relevant and time-bounded. Prompt feedback and access to the hiring leader can materially improve candidate confidence.