The hsh-finetune-dataset tool is live in the DOD catalog. Your agent describes the task, pays per-tier in USDC via x402, and receives an answer-verified, HuggingFace-ready dataset. Same product as the human page — callable end to end.
HSH·Intelligencehsh-finetune-dataset toolDiscover it via MCP at the endpoint above, or hit the x402 resource directly. The tool is live now and listed in the DOD catalog. Describe the task in plain language — what the model should learn to do — plus the row count and, optionally, the platform you'll train on.
POST /x402/t/hsh-finetune-dataset
{
"task_description": "step-by-step math word-problem reasoning with verified answers",
"domain": "math",
"row_count": 3000,
"target_platform": "gradients",
"verification": "programmatic"
}Pricing is by row count: 1–2K rows $75 · 2–5K $150 · 5–10K $300 (larger scoped on request). Settle over x402 on Base mainnet or Solana — the catalog entry carries both payment rails. No card, no account, no form.
The tool returns an order reference and a track URL. The dataset is built and answer-verified — every checkable answer is confirmed in code, deduplicated, and split train/val/test. Delivery is async; poll the track URL for status.
{
"ok": true,
"order": {
"reference": "HSH-XXXXXXXX",
"track_url": "https://dod.hshintelligence.com/track/HSH-XXXXXXXX"
}
}Delivered in Alpaca schema (instruction / input / output) by default — drop it straight into Gradients (SN56), TRL, Axolotl, or Unsloth. See a live verified sample before you buy: