pbperbug
AI work marketplace

Human intelligence for artificial intelligence.

Perbug connects AI companies with people who evaluate, annotate, and improve AI. Do real AI work, build a public reputation, and get matched with higher-paying projects as you prove your accuracy.

prove what you're good at · get your Proof of Merit in ~3 min

evaluate
AI responses
annotate
any modality
red-team
find failures
reputation
that compounds

Proof of Merit — your work speaks for itself

A reputation that's measured, not claimed.

Anyone can say they're a great evaluator. Perbug measures it — a transparent Proof of Merit score per skill, built from your completed work, blind quality checks, and domain tests. Your profile at perbug.com/you becomes a portable, verifiable reputation AI companies can search.

  • Per-skill scores you can see the basis for — no black box
  • Higher merit unlocks higher-paying, private projects
  • You control what performance data is public
AX
Alextop 5%
perbug.com/alex
94.7
proof of merit
Reasoning
97
AI evaluation
98
Coding
96
Writing
93
Mathematics
91

Measured across 4,281 completed evaluations, blind quality checks, domain tests, consistency, and client feedback — not self-reported.

The work AI companies actually need

Frontier models are trained and graded by people. Perbug routes that work to qualified humans and returns clean, provenance-tracked results.

AI response evaluation

Compare two answers, pick the better one, explain why.

RLHF ranking & critique

Rank and critique model responses to align behavior.

Data annotation

Label text, images, audio, and video with gold-set QA.

AI coding evaluation

Review generated code for correctness and quality.

Search evaluation

Judge whether results actually answer the query.

Domain expert tasks

Legal, finance, medicine, engineering, mathematics.

Red teaming

Surface hallucinations, jailbreaks, and failures.

Freeform rewrite

Improve, correct, and rewrite model output.

for AI companies

Buy specialized human evaluation

Deposit a project budget. Perbug handles recruitment, qualification, task distribution, and quality control — and returns the completed evaluation or labeled dataset. You pay for results.

  • Search & recruit qualified experts by skill and accuracy
  • Gold-set / consensus / expert quality control built in
  • Provenance-tracked exports (JSONL / CSV / Parquet)
for workers

Get paid for AI expertise

Pass a qualification, get matched to projects that fit your skills and languages, and get paid per verified item. Payouts via PayPal.

  • Remote, flexible, priced by outcome not by the hour
  • Domain experts (math, law, medicine, code) earn premiums
  • Reputation unlocks private, higher-paying programs

From signup to higher-paying work

  1. 01
    Create profile
    Skills, languages, specializations.
  2. 02
    Qualify
    Pass a task-type qualification test.
  3. 03
    Get matched
    Projects that fit your expertise.
  4. 04
    Do the work
    Evaluate, annotate, red-team, rank.
  5. 05
    Build reputation
    Unlock premium projects & invites.

Do AI work worth putting your name on.

Build the reputation that gets you recruited — or hire the humans who make your models better.