Product Architecture

How ChemEng Student works

A modern learning platform that combines a curriculum-aware AI tutor, structured practice, and a skills-driven progression engine.

System Overview

Client

Student App

Dashboard, Lessons, Practice

Instructor Studio

Course Builder & Reviews

Admin Console

Users, Plans, Content

Application Layer

AI Tutor Service

Curriculum-aware LLM

Rules Engine

Progression & Access Control

Analytics Engine

Weak-topic detection

Data & Infra

PostgreSQL

Content, progress, users

Billing

Subscriptions & seats

Simulation Runtime

Interactive labs

Core Learning Flow

Every lesson follows a structured, adaptive loop.

1. Lesson Delivery

Video · Objectives · Equations · Common mistakes

Checkpoint

▼

2. Practice Questions

Hints → attempt → worked solution

Rubric scoring

▼

3. AI Tutor Support

Explain · Solve · Review · Test me

Mastery signal

▼

4. Skill Update

Level ↑ · weak-topic flagging · path recommendation

Rules Engine

Deterministic policies that gate content, unlock certifications, and personalise recommendations.

RULE · access

Plan-based content unlock

IF user.plan == "free"
  AND lesson.tier > "preview"
THEN require_upgrade(plan="plus")

RULE · progression

Lesson completion criteria

IF video.watched >= 0.85
  AND checkpoint.score >= 0.7
THEN mark_complete(lesson)
  AND emit(skill_xp: +12)

RULE · weak topic

Adaptive recommendation

FOR skill IN user.skills
  IF mastery(skill) < 0.5
    AND attempts(skill) > 3
  THEN queue_practice(skill, count=5)

RULE · certification

Skill certification

IF skill.mastery >= 0.85
  AND assessment.passed == true
  AND lessons_covered >= module.required
THEN issue_certificate(skill)

AI Tutor Pipeline

Every AI response is grounded in curriculum-approved content.

STAGE 1

User Query

Lesson context, mode & message

STAGE 2

Retrieval

Fetch approved lesson notes, examples, rubrics

STAGE 3

Rerank

Score chunks against curriculum weights

STAGE 4

LLM Generation

Contextual response with citations

STAGE 5

Guardrails

Safety, rubric alignment, source check

Data Model (Excerpt)

Row-level security on every user-scoped table. Admin roles stored in a separate table.

profiles

· id
· email
· display_name
· plan_id
· created_at

subjects

· id
· name
· slug
· sort_order

lessons

· id
· module_id
· title
· video_url
· objectives[]
· equations[]

questions

· id
· lesson_id
· prompt
· difficulty
· skill_id

lesson_progress

· user_id
· lesson_id
· progress
· completed_at

question_attempts

· user_id
· question_id
· answer
· score
· attempted_at

user_skill_progress

· user_id
· skill_id
· level
· mastery

ai_conversations

· id
· user_id
· mode
· lesson_id
· created_at