For AI Teams

Human Data for Training,
Annotation and Evaluation

Collect speech, images, video and text β€” or label and score model outputs β€” with contributors in 200+ countries. Self-serve to start; HITL API when the product needs people in the loop.

Contributors in 200+ countries ready to start

200+

Countries

195,000+

Contributors available

~4 Hours

Avg time to results

AI Paths

Collect, Label, Score
or Call Humans Live

Same review gate for every path: contributors work once, you approve, then they get paid. Pick the path that matches the gap in your pipeline.

How It Works

Brief In.
Approved Data Out.

No enterprise onboarding to start. Launch a small test, tighten the brief, then extend the same task.

Create an AI task
  1. Pick the work type

    Collection, annotation or evaluation. Write the brief, examples, pay and how many submissions you need.

  2. Lock the geo

    Choose countries. Only adults there see the task β€” languages and demographics live in your instructions.

  3. One person, one pass

    Contributors capture, label or rate once. Duplicate attempts from the same person do not stack.

  4. Approve into the set

    Review in the dashboard. What you reject never enters the dataset and is refunded.

Who uses it

Built for teams shipping models

If you need human data or judgments without a six-week vendor cycle β€” start with a pilot task.

S

AI startups

Ship a training or eval set without waiting on vendor onboarding.

M

ML engineers

Collect speech, image or text that matches the users your model will see.

P

Product AI teams

Score model answers and plug human review into live product flows.

R

Applied research labs

Run preference and quality studies across many countries on one brief.

D

Data ops

Scale annotation with clear instructions and a reject-before-payout gate.

A

Platform / HITL owners

Call people from the product when the model is unsure β€” via API.

Strategic Partnerships

Open to partnerships, investment and acquisition

Need volume pipelines, custom QC or a closer data partnership β€” talk to us.

  • Collection, annotation, evaluation and HITL on one platform.
  • Geo targeting and review before every payout.
  • 195,000+ contributors across 200+ countries.
  • Confidential talks on integrations, ventures and acquisitions.
Explore partnerships
Partners in conversation over laptops during a working meeting
AI team FAQs

Frequently Asked Questions

Four things: collect new training data, annotate content you already have, evaluate model outputs with independent raters, or call people live from your product through the Human-in-the-Loop API.

Collection creates new files: recordings, photos, clips or writing. Annotation labels or rates content you already own, and evaluation has many raters score model outputs on one rubric.

Yes. Pick countries on the task, and only adults there see it. State the language and demographics in your instructions; contributors in 200+ countries can take part.

You set pay from $0.05 per task plus a 15% fee. There's no contract, subscription or minimum volume, so you can run a pilot and then extend it.

Each contributor works on your task once, so duplicate attempts don't stack. You review every submission, and what you reject never enters the dataset and is refunded.

Average time to results is about 4 hours, with 195,000+ contributors available.

Yes. The Human-in-the-Loop API returns live human decisions and labels to your product, and developer accounts can automate campaigns through the API.

Launch Your First AI Task

Start with a small collection, annotation or eval pilot β€” then scale the same brief.