ExamAnalyticsAI.

The problem

An assessment platform needs to connect question-level attempts with useful follow-up: what needs attention, which records should a parent or educator review, and what practice comes next. ExamAnalyticsAI implements services for these related workflows.

Implemented academic workflows

• Academic-health computation: a Python service computes an academic-health result and stores snapshots for later retrieval. These are software-generated indicators, not independently validated predictions of exam performance.
• Early-warning records: a service evaluates alerts, lists active alerts and marks them read. The inspected implementation provides detection and a read surface; email, push and SMS delivery are described as later work.
• Mistake diagnosis: a diagnosis service includes a heuristic fallback when AI-based analysis is unavailable. Its output quality and educational suitability have not been independently evaluated.
• Test generation: a service builds test blueprints and question sets. Curriculum alignment is an implementation intent; it is not evidence that every generated question meets a curriculum or assessment standard.

Architecture

The source combines a Next.js and TypeScript frontend with asynchronous Python FastAPI services, SQLAlchemy, asyncpg and PostgreSQL. Separate backend services handle academic-health calculation, warning records, mistake diagnosis and test generation.

The portfolio's Cloudflare routing Worker selects distinct upstream hostnames for development and production. Routing configuration is separate from confirmation that the upstream application is healthy.

Engineering boundaries

Persisted snapshots and alert read states separate analysis from the review workflow. A heuristic diagnosis fallback provides an alternative path when AI analysis is unavailable. Neither mechanism establishes predictive accuracy or a measured learning benefit.

Delivery state and outcomes

The application source has been inspected. Its test suites and live student transactions were not run in this audit. The concrete result is an implemented set of assessment-analysis services; no student improvement, adoption, accuracy or response-time figure is claimed.

Application screenshots remain pending explicit permission. Any approved assessment example will use synthetic student data.

Technology and classification

Product implementation, shown separately from direct client engagements and professional-peer testimonials.