AssessReview AI (ARAI) is a browser-based, AI-powered quality assurance tool developed for the College of Pharmacy at Gulf Medical University. It automates the peer-review process for ExamSoft examination files, applying 21 standardised assessment-quality criteria aligned with the QF Emirates National Qualifications Framework.

Unlike a manual checklist, ARAI re-evaluates its findings in real time as you upload different files. This allows reviewers to check alternative versions and understand exactly why a defect is reported.

​

Key Features:

• 21-criterion AI analysis: Checks language clarity, Bloom's taxonomy alignment, blueprint coverage, option design, clinical accuracy, and more.

• Direct PDF reading for CLO/PLO/Bloom's: Sends the exam PDF natively to Gemini to read the Item Categories table — eliminating false positives from text extraction.

• Word-by-word similarity detection: Jaccard similarity at ≥ 95% threshold flags reused items against previous exams, reporting the exact Item ID of the match.

• Model cascade: Automatically tries up to 8 Gemini models in order — if one is overloaded it falls back to the next without user intervention.

• Structured Word export: Produces a formatted Assessment Review Report (.docx) that exactly mirrors the on-screen layout.

• Session restore: Completed reviews are saved in the browser and can be restored on the same device.

• Zero server storage: All processing happens in the user's browser — no exam data is stored on any server.

• Responsive design: Fully functional on desktop, tablet, and mobile screens.

​

Professional review required

The ARAI Defect Report is a structured decision-support output, not a validated psychometric instrument. All AI-generated defects must be reviewed and confirmed by a qualified academic before any action is taken.

​

How it works?

See the full documentation file