For AI teams

Human expertise for your training data.

Tell us what your model needs to learn. We put together reviewed people with the right background, starting with a small pilot so you can judge the quality first.

Tell us about your project See what we can do

What we can help with

Each project is scoped to your task, and we'll tell you up front if we're not the right fit.

Response evaluation

People compare and rank model outputs against your guidelines, and explain why.

Prompt and response writing

Original prompts and model answers written by people who know the subject.

Domain-expert review

Accuracy checks by people with a background in fields like medicine, life sciences and law.

Code evaluation

Programmers solve problems, review generated code and find where it fails.

Bangla and multilingual data

Writing, review and evaluation by people who work in the language every day.

Adversarial testing

Structured attempts to find where your model gives unsafe or wrong answers.

How we work

  1. Scope

    We define the task, skills, volume and what good work looks like, using your guidelines.

  2. Assemble

    We select people from our reviewed community and ask each one before they join your project.

  3. Pilot

    A small batch comes first. You review it and we adjust before committing to more.

  4. Deliver

    We scale what works, and review work against your guidelines as we go.

Common questions

Who would work on our project?

People from the Tabiz community, whose profiles we've reviewed by hand. We select for your task and share relevant background with you once the person has agreed.

How is our data handled?

We agree how your data and work will be handled before anything is shared, and can work under an NDA.

What does it cost?

Pricing depends on the expertise, volume and complexity of the task, so we quote each project once it's scoped. A pilot lets you test the work on a small budget first.

What's the smallest project you take?

Tell us what you have in mind. We'd rather start small and prove the quality than promise too much.

Tell us about your project

We read every inquiry and reply by email. Please don't include confidential data here.