Skip to main content
8 task types · 1 API

What workers do for you

Real humans complete your tasks inside the SLA you set. Every response comes back as typed JSON. Chain them with LLM / parse / exec primitives via pipelines when you need hybrid workflows.

🏷️

Training data labeling

Build supervised datasets: sentiment, intent, NER, toxicity, bounding boxes. Run 3–20 workers per task with majority vote for consensus.

label_textlabel_image
🛡️

Content moderation

Route user-reported posts to humans for final approve/reject decisions with a written reason. Pair with policy context for consistent rulings.

moderate_content

AI output quality review

Have workers rate LLM-generated summaries, answers, or images on a 1–5 scale with written critique. Use the results as an eval set or RLHF signal.

rate_qualitycompare_rank

Fact verification

Check claims against sources. Workers return true / false / unverifiable with a citation. Useful for news aggregators, knowledge bases, and LLM output grounding.

verify_fact
📝

Transcription QA

Send an AI-generated transcript to a worker for correction. Combine with an LLM pre-pass in a pipeline so humans only touch segments flagged as uncertain.

transcription_review
💬

Domain Q&A and research

Ask workers to answer free-form questions with optional context. Route by skill certification so specialists get specialist tasks.

answer_question

Have a different workflow?

Tell us what you want to build. We're onboarding beta users one at a time.