Edusify

The Problem

Generic quiz apps can't render the exam - or trust the bank. Mixed item types like hotspot, drag-and-drop, case study, active screen, and simulated labs go unsupported by most quiz tools. Question banks are often leaked or brain-dumped, undermining trust and value. Learners rarely get a per-domain diagnosis of where they're weak, and generic MCQ tools give no feel for the navigation and chrome of a real exam under time pressure.

Objectives

• Render all nine real Microsoft-style interactive item types, not just multiple choice.
• Give every learner a scaled, domain-weighted score and a clear view of their weak objectives.
• Keep the question bank trustworthy: every item cites an official source and passes human review before publishing.

Solution Design

We built a Next.js learner and admin product on top of a dedicated FastAPI backend, with SQLite as the volume-backed store - no external database, no Redis or Celery.

Every session - study or exam - samples from published questions only, grades each of the nine interactive item types against a stored correct payload, and computes a scaled estimate from the pack's own min/max and passing score, with domain-weighted sampling that mirrors real exam structure. New certification packs ship as data files under exam_packs/{code} - YAML metadata, objectives, and seed questions - with no engine changes required. AI-assisted generation drafts new questions against strict JSON schemas, checks them for duplication, scores their quality, and holds them as pending until an admin accepts them into the live pool.

Implementation Process

Results & Business Outcomes