Send us a real task. We deliver 200 to 500 evaluated or labeled items, with a written quality report, at a fixed price agreed before we start.
We reply within one business day.
Five steps from your first message to delivered data.
We agree the task, sample data, guidelines, accuracy target, volume and a fixed price.
We sign your NDA or ours and a short order form. You share sample data securely.
We label about 25 items, review them with you and tighten the guidelines.
Specialists complete the pilot; a second reviewer checks every batch against hidden test items.
You receive the files and quality report, and we walk you through the results.
One secure delivery package, in formats your team can use straight away.
| File | What it is |
|---|---|
| data.jsonl | The pilot items, one record per line, ready for training or evaluation code |
| data.csv | The same items as a spreadsheet for human review |
| data_card.md | Fields, method, counts, known limits and version |
| quality_report.pdf | Accuracy, agreement scores and error examples |
| guidelines.pdf | The final labeling or evaluation guidelines |
| SHA256SUMS.txt | Checksums so you can confirm files arrived intact |
Delivered to your S3, Google Cloud or Azure bucket, a private Hugging Face repository, SFTP, or a secure link that expires after 7 days. We can also work directly inside your labeling platform.
{"id": "FIN-EVAL-0001",
"domain": "financial_accounting",
"topic": "Prepaid expenses / adjusting entries",
"difficulty": "foundation",
"question": "On 1 January a company pays $12,000 for a 12-month insurance policy and records it as prepaid insurance. What adjusting entry is needed on 31 January?",
"reference_answer": "Debit Insurance Expense $1,000; credit Prepaid Insurance $1,000.",
"reasoning": "$12,000 / 12 months = $1,000 of the policy used in January.",
"rubric": ["Correct accounts", "Correct amount ($1,000)", "Debit/credit direction correct"],
"reviewer_tier": "senior_finance_reviewer",
"qa_status": "approved"}
Every pilot is fixed-price and scoped on the call. Typical examples:
Up to 500 graded finance questions with rubrics, or 300 of your AI's answers checked and corrected.
Up to 500 preference rankings or rated responses, with reasons for each decision.
Up to 300 Arabic or bilingual responses rated by native speakers.
Up to 500 documents, texts or images labeled to your schema.
Pilots are fixed-price and depend on the task and volume. We confirm the price on the scoping call, before any work starts.
Projects are contracted by Vital Hawk Inventions in the UAE and delivered by our group partner Yuztech's managed teams, mainly in Pakistan. Only the named team members on your project can access your data, and we confirm the processing locations in writing.
Yes. We can work in your platform or set up Label Studio or CVAT for you.
If you are happy with the results, we agree a production plan: per-item pricing, a monthly team, or a fixed-scope project. There is no obligation to continue.
We run small paid pilots so both sides commit real time and you get production-quality work, not a sales demo.
Tell us about your task. We reply within one business day.