Your first AI hire should remove a specific constraint. “We need someone to lead AI” is a starting conversation, but it leaves candidates guessing about the job and interviewers guessing about the hiring bar. Before opening the search, identify the next product or technical milestone that the current team cannot deliver.
Start with the work already waiting
Write down three concrete tasks the person would tackle. They might need to build an evaluation set, integrate a model into a product workflow, or diagnose an inference bottleneck. Include the context: what exists today, who will support the work and what the business needs to learn.
Then ask whether the obstacle is engineering capacity, specialist knowledge or uncertainty about the product. If nobody knows what customers need, hiring a researcher will not resolve the product question by itself. If the product direction is clear but deployment is fragile, additional research depth may not be the immediate priority.
Choose a centre of gravity
- Applied AI engineer: consider this profile when the main job is building useful product behaviour around models, data and evaluations.
- ML engineer: consider this profile when the core work includes model pipelines, experimentation and dependable operation.
- Research engineer or scientist: consider this profile when progress depends on developing or testing genuinely uncertain modelling approaches.
Titles overlap between employers. Treat these as descriptions to calibrate, not universal definitions. Ask candidates about actual ownership rather than assuming their previous title describes their capabilities.
Account for the team around the hire
A specialist can be effective inside an experienced engineering organisation and struggle in an unsupported first-hire role. State who owns the product, infrastructure, data access and technical review. If the answer to every question is “the new person,” narrow the scope or plan additional support.
For example, a small team building a document assistant might need someone comfortable with backend engineering and evaluation. A team investigating a new model architecture could need a very different background. The word “AI” does not make those briefs interchangeable.
Turn the choice into interview evidence
Give the candidate a realistic scenario based on the first milestone. Ask what they would clarify, how they would reduce uncertainty and what they would deliver first. Look for a coherent plan that fits your constraints. Be cautious when a solution assumes a larger team or cleaner data than you actually have.
Use a past-project discussion to check whether they have operated with similar ambiguity. Separate their own decisions from the team’s output, and allow anonymised examples.
A focused role creates a more honest proposition for candidates and a clearer decision for your team. Use the AI engineer hiring guide to build the assessment around that decision.