The Expert Index Score: How We Measure AI Talent Objectively

A resume tells you where someone has worked. It does not tell you whether they can actually fine-tune a model, debug a broken RAG pipeline at 2am, or spot a subtly biased reward signal before it ships. For AI-native roles, the gap between "looks qualified on paper" and "can do the work" is wider than in almost any other field — and it is the reason we built the Expert Index Score instead of relying on resumes and reference checks alone.

Every expert on the platform is scored the same way: a composite of three independent signals, each measuring something the others cannot.

Interview performance (40%) — a structured, AI-conducted technical interview scoped to the expert's domain. Unlike a generic behavioral screen, questions adapt in real time based on the answers given, probing deeper wherever a response suggests either strong or shallow understanding.

Resume intelligence (30%) — an NLP engine parses work history, projects, and publications to score depth and domain relevance. This catches signal a keyword search misses: the difference between "used PyTorch" and "shipped three production PyTorch models," for instance.

Micro-task performance (30%) — three short, live-scored challenges spanning coding, reasoning, and debugging. These are the closest thing to watching someone actually work, and they are the hardest component to game.

The score is not a one-time certificate. Every completed task, every client review, and every new interview attempt feeds back into it, so a score from six months ago and a score from this week can tell meaningfully different stories. That is by design — capability compounds, and so should the credential that represents it.

A high score does not just look good on a profile. It is the primary signal Hire Hub uses when matching experts to incoming requests, and it determines which opportunities surface first. The mechanics are transparent on purpose: if you know what is being measured, you know exactly what to work on next.