AQR KPI2.1 AI is a browser-based quality assurance tool built for the College of Pharmacy at Gulf Medical University. It supports the internal assessment review carried out by the Assessment Committee against CAA OBEF KPI 2.1, producing a scored, evidence-referenced dossier for each course under review.

The application evaluates every assessment element of a course against nine rubric criteria grouped into four weighted domains, and computes a composite score and a KPI percentage. Each criterion score is accompanied by the evidence that produced it, what was found, what was missing, and — where relevant — the exact measurement behind the judgement.

Two engines are available. The local rule engine applies deterministic, auditable rules entirely in the browser with no network calls. The AI engine additionally sends de-identified evidence to Google Gemini for criteria that require reading and interpretation. Criterion 2c is always deterministic and cannot be overridden by the model.

​

Key Features:

• Nine-criterion OBEF rubric across four weighted domains: alignment with learning outcomes, assessment design, grading clarity, and feedback — producing a composite of 1.00–5.00 and a KPI of 20–100%.

• Genuine current-versus-previous comparison for item reuse: questions are extracted from both papers and matched question by question, with whole-document comparison as a second basis for assessments that are not question papers.

• Automatic model selection: the Gemini model catalogue is ranked live and the newest stable model is chosen, with automatic step-down through up to eight candidates when one is unavailable.

• Browser-side de-identification: student names, identifiers, emails and phone numbers are stripped from grade files at upload, before any analysis and before any network call.

• PDF export: that is an exact copy of the on-screen report, stamped PROVISIONAL until the review is signed off.

​

Professional review required

The scores and defects produced by this application are structured decision support, not a validated psychometric instrument. Every criterion score and every reported defect must be reviewed and confirmed by a qualified academic before it informs any decision or is submitted to the Assessment Office.

​

How it works?

See the full documentation file