NKODATALABS/medical-device-safety-reasoning-preview

Medical Device Safety Reasoning — Private Preview

NKO Data Labs · commercial dataset preview · target commercial release: 100k+ curated records

A structured safety-reasoning dataset derived from public FDA/openFDA device adverse-event records for model evaluation, classification and grounded safety analysis.

Preview status

This repository is a small authentic preview, not the completed commercial corpus. The full commercial release is under development and will only be produced if market demand is validated.

Intended ML tasks

• —adverse-event narrative classification
• —device-problem / outcome extraction
• —structured JSON generation
• —evidence-grounded safety summarization
• —temporal evaluation

Preview fields

mdr_report_key, event_key, dates, event type, manufacturer/device metadata, product code, patient outcome codes, narratives, source/provenance.

Provenance

Source family: FDA MAUDE data exposed through openFDA. Preview preserves source identifiers and provenance. Passive safety reports can be incomplete or unverified and must not be used to estimate incidence or clinical causality.

Delivery target

Commercial release target: JSONL + Parquet, documented schema, QA report, temporal train/validation/test split and optional task-specific labels.

Commercial access

Indicative commercial licensing starts at USD 2,500 for non-exclusive internal use; enterprise/model-training terms depend on scope. Request access to discuss required volume, use case and delivery format.

Limitations

This preview is for product evaluation and market validation. No claim is made that the target 100k+ commercial corpus is currently available.