We build a clean, answer-verified dataset and deliver it HuggingFace-ready. Drop the repo straight into Gradients, TRL, Axolotl, or Unsloth. Every checkable answer is verified in code, not trusted from a model.
train split
ground truth 120.75computed in Python, not the model
answer matches ground truth, row kept
Most synthetic datasets trust the language model to be right. This one does not. The correct answer is computed independently before the model writes a word, then the model’s answer is checked against it.
step 01
Each problem is built in code with a known-correct answer as ground truth.
step 02
The model writes step-by-step reasoning and a final answer for the problem.
step 03
The model's answer is checked against ground truth in code. Mismatches are discarded.
step 04
Deduplicated, split train/val/test, documented, pushed to a HuggingFace repo.
100%
Alpaca
24h
Agents discover and buy this dataset programmatically over MCP. Describe the task, get a quote, pay per-call in USDC via x402. No card, no form.
Order
Four published tiers. Change the task, domain, or rows and the live price updates. Specialized work is quoted in 24 hours.
01
Choose the job
Task, domain, rows, format.
02
Price locks
List price, or a 24-hour quote.
03
We deliver
Answer-verified HuggingFace repo.
Four published builds, priced as a job. Change the order form and the live price follows the tier, unless the work is specialized, in which case it is quoted.
Narrow-task LoRA
for first-time fine-tuners
$75 /build
Select planDomain LoRA, staging-ready
for startups and ML engineers
$150 /build
Select planProduction domain model
for teams that need to trust it
$300 /build
Select planDeep domain specialist
for labs that need depth
$600 /build
Select plan25,000+ rows, or a shape that does not fit a tier.
Multi-language, red-team, multi-modal, large corpora.
Price on request /build
Select plan24-hour scope review
After payment, every order gets a scope review within 24 hours. If the request needs domain expertise, sources, or verification depth beyond the quoted price, we propose an adjusted scope first.
Full-refund fallback
If no acceptable adjustment exists, you get a full refund within 2 business days. Revisions for errors in the delivered dataset are always included. Changes to the task definition itself are a new order.
What we do not build
HIPAA-regulated patient data, classified or restricted-source material, real-time data streams, or expert credentials we do not hold. These are flagged at scope review and automatically refunded.
Live
HSH Intelligence
Any data you can describe, built to order, and callable by a human or an agent.
HSH Intelligence
Skip to content
