About this role
Role Overview
- Protect dataset quality at scale by running sampling-based audits, enforcing quality gates, and ensuring every dataset release is audit-ready and meets acceptance thresholds.
- Execute sampling plans (stratified by label type, difficulty, site/surgeon/device, model confidence) for ongoing QA and release audits.
- Audit annotations against guidelines + ontology; score defects using standardized rubrics (severity, type, root cause).
- Run and validate quality gates prior to release: completeness/coverage checks schema/constraint checks (illegal combinations, temporal consistency, boundary rules) defect density and rework thresholds.
- Generate rework tickets with precise instructions; track closure and verify fixes.
- Partner with Annotation Lead to drive calibration actions (training refreshers, guideline clarifications, examples library).
- Maintain QA dashboards: defect rate, rework %, gate pass rate, audit cycle time, top error patterns by annotator/team/label type.
- Support release readiness: ensure required artifacts are complete for the Evidence Pack (QA results, sampling logs, sign-offs).
- Escalate systematic issues to Adjudication/Clinical Review (definition gaps, ambiguous classes, recurring edge cases).
Requirements
- 3–7+ years in data QA / annotation QA / operations quality; experience auditing labeled datasets (video preferred).
- Strong command of sampling-based QA: stratified sampling, acceptance thresholds, and audit scoring.
- High attention to detail; comfortable reading and enforcing detailed SOPs/guidelines.
- Strong documentation skills; can write clear defect descriptions and rework instructions.
- Comfortable with dashboards/spreadsheets; basic metric literacy (defect density, rework, throughput).