AI & Data

We ask for one specific model or system in live use, its evaluation approach, and what degraded over time. Notebook work and production experience look identical on paper.

For LLM roles we probe how they measured quality beyond vibes — datasets, offline evaluation, human review, regression detection on prompt or model changes.

In credit, fraud and financial decisioning, a model that cannot be explained cannot be deployed. We screen for candidates who have worked inside that constraint.

Writing a research brief when the need is someone to productionise and operate models.

Hiring AI talent before the data platform can support them — a reliable cause of attrition.

Screening LLM engineers on modelling theory rather than evaluation and cost control.