← Back to Work
Project · Product Build

AI Teacher — Interactive Math & AI Learning Platform

A from-scratch learning platform with 27 courses, from College Algebra through Deep Learning, built around 548 interactive widgets and a pluggable AI tutor.

Focus: content architecture at scale, custom interactive UI, symbolic answer checking, and AI used as a teaching tool.

Courses
27
Interactive widgets
548
Back end
FastAPI + PostgreSQL

Try the live demo ↗

What it is

Hundreds of lessons across College Algebra, Calculus, Linear Algebra, Probability, Deep Learning, and Language Models. Every concept has a rigorous Textbook lane and a Plain English translation, plus an optional AI explanation lane and a Socratic AI tutor.

The AI features are provider-pluggable: Anthropic, OpenRouter, Ollama, or LM Studio. Answers are checked symbolically with SymPy on the server, and mastery is tracked per skill.

Screenshots

Stage 1

Curriculum Breadth & Course Tree

A published course catalog spanning core math through applied AI, and a chapter-by-chapter course tree with per-section time estimates and progress tracking.

Stage 2

Dual Explanations, Light & Dark

Every concept is written twice: a rigorous Textbook lane and a genuine Plain English translation for a reader starting cold — shown here in both of the platform's light and dark themes.

Stage 3

548 Interactive Widgets, In Context

Widgets are embedded directly in the lesson, not off in a gallery — 2D JSXGraph explorers, algorithm-style scenes, and fully rotatable 3D scenes built on three.js, one per idea that benefits from being dragged instead of just read.

Stage 4

Built-in Study Tools

A notes-and-highlighting system anchored to the exact passage selected, and a Desmos-style graphing scratchpad available on every page — no copy-pasting into a separate tool.

Stage 5

Curriculum & Skill Dependency Graphs

The entire 27-course curriculum as one derived dependency graph, down to a single concept's own local prerequisites — every edge computed from the content itself, never hand-maintained.

Stack

FastAPI with Jinja2 server rendering, HTMX, Alpine, PostgreSQL (raw SQL), KaTeX, JSXGraph, three.js, SymPy. Content is authored as version-controlled Markdown.

Live demo: aiteacher.fly.dev