STEP 01
WHAT WE FLAG
Catch the issue
Rules catch the obvious. Rubric checks surface the harder ones. Severity, confidence, and rationale stay with the case.
→MODELS · AURAQC
Quality that doesn't end at ship day.
A score does not survive an audit. A reviewed case does. The flagged case keeps its reviewer, its rationale, and its decision from first flag to final call.
ISSUE
One queue
Every flagged case lands in the same place.
REVIEW
Right reviewer
Routed by the rubric, not the rota.
RECORD
Closed with proof
Decision, evidence, and reviewer stay attached.
COVERAGE READING
ISSUE TYPE × REVIEWER LANE
HOW IT WORKS
Three steps. One record.
Catch the issue. Route it to the right reviewer. Close it with the record included.
STEP 02
WHO WE GIVE IT TO
Route to the right reviewer
The case lands with the person who knows this rubric. They see the artifact, the history, and the rule that fired.
→STEP 03
WHAT WE KEEP
Close with the record included
The decision, the reviewer, and the evidence stay on the record. Reopened later, the case still reads the same.
WHAT COMES OUT
What the queue leaves behind.
Every flagged case leaves something the team can act on — and something the next reviewer can read.
01
Scorecards
One read on what was flagged, what was reviewed, and what the call was.
↳ ARTIFACT
02
Reviewer queues
Cases routed to the right reviewer with the rubric reading and artifact attached.
↳ ARTIFACT
03
Override timelines
What changed. Who decided. What the rubric said before and after the call.
↳ ARTIFACT
04
Review records
Rule, artifact, reviewer note, and verdict — ready when someone asks.
↳ ARTIFACT
WHERE IT FITS
In the loop, this is where you review.
Test the run. Review the hard cases. Recruit the right specialist. Remember the misses. Approve what's right.
01
Test
02
Review
● YOU ARE HERE
03
Recruit
04
Remember
05
Approve
FAQ · WHAT TEAMS ASK
Five common questions, direct answers.
Q · 01
How is this different from our monitoring dashboard?
A dashboard tells you what changed in production. AuraQC records who reviewed it and what they decided. A flag becomes a reviewer assignment, a rationale, and a closed verdict — not a chart and a Slack thread.
Q · 02
Do we keep the model we improve on this work?
The reviewed cases, decisions, packet evidence, and handoff terms stay attached to your record. Model export or tuning terms are scoped in the engagement rather than claimed by default.
Q · 03
Can we point it at our own models and queues?
Yes. AuraQC is model-agnostic — frontier APIs, on-prem checkpoints, and your own tuned models. It plugs into the review queue your team already runs.
Q · 04
Where do the runs execute, and who can see the cases?
Runs execute where your checkpoints live. Cases and reviewer records stay isolated to your program — not pooled, not shared with a single data vendor whose interests differ from yours.
Q · 05
Does this help us pass an audit?
The rule that fired, the artifact, the reviewer note, and the verdict stay attached to every case — the provenance a high-risk audit asks for under the August 2026 deadline.
RELATED MODULES
Next to this in Models.
EVALUATION STUDIO
Review it before release.
Write the rubric once. Run every release against it. Ship with the record included.
See the page →REGRESSION BANK
Every mistake. Only once.
Every escaped failure becomes a gate the next release cannot cross.
See the page →COMPLIANCE MONITORING
Compliance records without scramble.
The record builds as the work is done — ready for the August 2026 provenance deadline.
See the page →AURAQC
Quality that doesn't end at ship day.
Bring the review queue your team already runs. Every flagged case stays readable from first flag to final decision — and the proof stays attached when an auditor asks.