Back
QWardPREP
QWardPREP

Product Architecture

A composable, metadata-driven platform engineered for NGN simulation at scale.

System map
Client Apps
Flutter iOS · Android · Web
Edge API Gateway
Rate limits · JWT · request signing
NGN Engine
Rules · scoring · adaptivity
Analytics
Mastery · cohorts
Vimeo Pipeline
DRM lectures
Billing
Stripe · plans
Supabase · PostgreSQL
RLS · realtime · storage
Audit + Observability
Logs · SOC2-ready trails
NGN engine flow
  1. 1. Request next item
    Engine receives student state: mastery vector, recent errors, cadence.
  2. 2. Rule evaluation
    Metadata rules score candidates by category gap, difficulty fit, recency.
  3. 3. Adaptivity
    Item sampled probabilistically — never the same template twice in a session.
  4. 4. Scoring
    Polytomous scoring for SATA / Cloze / Hot Spot with partial credit.
  5. 5. Mastery update
    Bayesian update on category & sub-skill vectors; persisted to profile.
Items / week
3.8M
p95 latency
142ms
Uptime SLO
99.95%
Concurrent sessions
12k
Roadmap to CAT
  • · Phase 1 — Rules-based adaptive (current)
  • · Phase 2 — IRT calibration on 50k+ responses
  • · Phase 3 — Full Computerized Adaptive Testing engine