Case study // 01
AI University Learning Platform
Active learning frontend that turns course content into adaptive AI study sessions, practice exams, and instructor analytics — not passive video watching.
Role · Frontend architecture & product engineering
Overview
University courses often ship as PDFs and videos with weak engagement and little visibility into what students actually understand. This platform helps instructors author structured courses while students study through timed, AI-evaluated sessions with tutoring and progress analytics.
The case study focuses on the presentation and orchestration layer: a modular Next.js client consuming an external REST API for AI evaluation, persistence, and media.
Architecture
Thin App Router shell with fat feature modules. Proxy enforces auth/RBAC; services call a centralized fetch client; TanStack Query caches server state; Zustand holds auth and wizard drafts. Next route handlers cover logout and a Range-capable video proxy.
- Shell / module boundaries for scalable feature ownership
- Cookie JWT access + refresh with SSR/CSR fetch client
- Learning session state machine (timers, breaks, distraction)
- Design system via Radix + CVA + Tailwind
Features
Challenges → Solutions
Complex learning session state
Orchestrated evaluate → assist → tutor flow with exit guards and flush semantics
Multi-tenant roles across App Router
Edge proxy allowlists + layout-level role gating
Instructor content ops in spreadsheets
FortuneSheet study-mode authoring with Excel import
Auth refresh storms
Coalesced token refresh in a centralized fetch client
Tech stack
Honesty note
This case study covers frontend architecture and product engineering against an external API. Backend ownership (schema, LLM gateway, scoring) is separate and not claimed here.
Next