For AI labs

Review-gated supervision at scale

Plug into our REST API with bearer auth. Send cases to POST /api/lab/upload and pull JSONL from GET /api/lab/export when your rows are ready. Built to fit a normal MLOps pipeline.

How it works

  1. Upload. POST cases to the lab API with your bearer token.
  2. Annotate. Verified physicians complete structured tasks on each case.
  3. Export. GET quality-checked JSONL when rows are ready.
API integration

Send an Authorization: Bearer header on every request. We issue a LAB_API_SECRET for your team when a pilot starts.

Upload a case

POST /api/lab/upload
Authorization: Bearer YOUR_LAB_API_SECRET
Content-Type: application/json

{
  "external_case_id": "DEMO-001",
  "title": "SBS demo case",
  "payload": {
    "prompt": "Which response is clinically safer?",
    "response_a": "…",
    "response_b": "…"
  }
}

Export approved annotations

GET /api/lab/export
Authorization: Bearer YOUR_LAB_API_SECRET

# Optional: ?source_lab=partner-lab
# Optional: ?mark_exported=false  (preview only)

Deliverables & pilots

  • Pairwise SBS evaluation is live today. Imaging causal graphs are on the roadmap.
  • We usually start with a scoped pilot to agree on format, volume, and SLA.
  • Every export row includes quality metrics and time-on-task per annotator.
  • Coming soon: multi-rater subsets with Cohen's κ and Fleiss' κ IRR reports.

Security

  • Row-level security and lab-only export APIs
  • Confidential watermarks and a full audit log of access
  • Only quality-gated rows leave the platform via the export API
Contact team