A new kodiva.ai: people for AI, data for frontier models

We started Kodiva in 2024 with one problem: hiring good engineers took too long, and CVs said too little. So we built an AI interview that tests what people can actually do.

Two years on, the same interview vets far more than engineers. Doctors, lawyers, mathematicians, linguists and analysts now go through it. Today we’re relaunching kodiva.ai around what the company has become: one expert network, two ways to work with it.

Frontier Data: experts who train and test AI

AI labs and AI companies need people with real expertise to write answers, rank model outputs and explain why, build evaluation sets, and find where models fail. Generic labelling can’t do this work. It needs a physician who can spot a dangerous drug interaction, or a senior engineer who can tell a clean fix from a fragile one.

Frontier Data gives AI teams vetted experts by specialty, country and language. We handle recruiting, vetting, quality review and payments, so teams can focus on their models.

Talent: AI-native people for companies

Companies building with AI need engineers, operators and analysts who already work that way. Talent matches them with people from the same network, interviewed and scored before anyone reaches a shortlist.

Why one network

The hard part of both businesses is the same: knowing who is genuinely good. Our network now spans 40,000+ experts in 60+ countries, and every one of them has been through a structured AI interview in their field. That’s what lets us move fast without guessing.

What’s next

We’ll publish more about how we vet experts, what AI teams are asking for, and, soon, our own evaluations of how models perform on expert tasks.

If you’re building AI and need experts, talk to us. If you’re an expert who wants to help train AI and get paid well for it, find work.

Varun Aggarwal, Co-founder & CEO, from Pune and San Francisco