Realset Workspace

Expert judgment, captured as training data

The workspace where domain experts answer real tasks, record how they got there, pass independent review, and get paid. Every approved record exports as JSONL with provenance you can verify.

The anatomy of a good record

Illustrative example from finance. Each record carries the task, the expert's reasoning, and a reviewer's decision.

01 / Real-world task

Revenue grew. Did performance improve?

A fictional distributor reports 20% revenue growth while gross margin falls from 30% to 23%.

Financial analysis Evidence required

02 / Expert contribution

"Growth alone is not enough."

"Gross profit fell from 30 to 27.6 on a revenue index of 100 → 120. Investigate pricing, product mix and input costs before attributing the change." Answer, supporting evidence, and a screen recording of the work.

03 / Independent review

Correctness · Evidence · Completeness

A separate reviewer checks the answer against the rubric, watches the recording when present, and records a decision. Only approved records are exported.

Quality gate

The method

From domain knowledge to dependable datasets. Start with a real task, keep the expertise, the evidence and the working session together.

01

Define the work

Set the question, source material, language and region, reward, intended data use, and an explicit evaluation rubric.

02

Capture expertise

Experts contribute an answer, supporting evidence, and, when the task calls for it, a screen recording of how they got there.

03

Review independently

A separate reviewer checks quality against the rubric, watches the recording when present, and records a decision.

04

Deliver and settle

Export approved records as JSONL with provenance and review metadata, while approved work accrues to the contributor's payout ledger.

Built around the work

For knowledge that doesn't fit a checkbox. First workflows cover finance and business operations; use synthetic examples to explore before introducing authorized enterprise data.

Global by default

Expert networks are global. Three things decide whether an expert outside your head office can actually work with you: language, payment, and trust without a meeting.

i18n

Localised, not just translated

Six interface languages including full right-to-left layout for Arabic. Every task carries an ISO 639-1 language and ISO 3166-1 region, so a dataset's coverage is a fact you can query. English · Simplified Chinese · Japanese · Korean · Spanish · Arabic

₮

Paid the way they want to be paid

USDT on TRON, Ethereum or BNB Smart Chain, USDC, BTC, IBAN and local bank transfer, PayPal, Wise, Alipay. Destinations are format-checked before they are stored, and every approved task accrues to a ledger with its own reconciliation export.

⏺

Provenance you can watch

Experts can record their screen while they work, directly in the browser. The recording is stored with a SHA-256 digest, shown to the reviewer alongside the answer, and referenced by hash in the export. WebM · SHA-256 · Reviewer playback

What the workspace covers today

A focused workspace. A complete data workflow.

In scopeTask creation, expert contributions with optional screen recording, independent review, approved JSONL export, payout ledger
IsolationEach workspace is separate: its tasks, recordings and ledger are visible only to the people invited into it
LedgerRecords what is owed and what an operator has settled; it does not move money, hold funds, or connect to an exchange
Not includedAutomated model evaluation, fine-tuning, enterprise SSO
AccessSign in with your operator-provided account, or ask us for a workspace
Built onLabel Studio Community Edition (Apache-2.0) for annotation storage

Tell us what your model needs

Three questions. We reply within one business day with a sample set and a quote.

Looking for capture or annotation work? Go to the Experts page instead.

Thanks, you're all set.

Your mail client opened with the request pre-filled. Send it and we'll reply within one business day.