Product
AuraQC keeps quality calm.
Rules, patterns, and ML work together to catch low-quality work early. Escalations arrive with evidence, not accusations.
Structural validation and rubric rules catch obvious violations with explainable evidence.
A lightweight classifier scores risk and routes edge cases into review instead of guessing.
High-confidence flags create queue-ready items with reason, evidence, and severity.
Reviewer feedback is captured and can be retrained on a monthly cadence when volume is sufficient.
Schema, types, required fields
Guidelines and rule validation
Too-fast, identical, low-effort
Risk scoring + triage
Queue assignment + evidence
AuraQC components, APIs, and dashboard exist in the repo. Precision/recall targets and time-savings metrics require production validation.
Every flag includes a reason and evidence payload so reviewers can decide quickly.
Structural, rubric, and pattern checks run before ML so obvious issues never wait.
High-confidence detections route into adjudication queues with severity and timestamps.