[01]Use case · Post-training

Teach the model what a correct answer actually looks like.

Supervised demonstrations and preference pairs written by people who do the work for a living — so the model learns the shape of a good answer, not the shape of the average one.

Deliverable
SFT demos + preference pairs
Priced by
per task
Credential gate
Practitioners, 5+ yrs
Typical rate
$60–140

What you're buying

A set of tasks with expert-authored demonstrations, plus preference pairs that rank a strong answer against a plausible-but-wrong one. Every task is named, scoped, and priced by the unit — no open-ended 'annotation hours.'

Who authors it

Practitioners with five or more years in the domain. The credential that qualifies them to write the answer is verified and attached to the artifact, so you can see exactly whose judgment trained your model.

How it's graded

Each task passes through a second credentialed reviewer before it ships. You get an inter-annotator agreement report alongside the data, not a promise that it was checked.

By the numbers
5+ yrs
author experience
review passes
A/B
preference pairs
0.9+
target agreement