%

EdTech

Case study

FAZERLANE*

Gamified learning platform that transforms YouTube tutorials into structured, AI-powered learning lanes with progress tracking and semantic content recommendations.

Project

FAZERLANE

Industry

EdTech

Year

2025 - 2026

Role

Software Engineer

The problem

YouTube is full of good tutorials, but watching one isn't the same as learning from it. There's no structure, nothing to check you actually understood, and no record of how far you've come.

The idea

Fazerlane turns a YouTube video into a lane: a set of challenges, in the style of Duolingo or Brilliant, generated from the video itself. You work through them, get graded, and earn XP and keep a streak going as you learn.

How it works

  • Paste a link and optionally pick a start and end time. The lane is created straight away and the challenges are generated in the background on a job queue.
  • Long videos are split into segments of up to 30 minutes. Gemini watches each segment and writes challenges for it, and it's shown the challenges it already wrote so it doesn't repeat them.
  • Generation progress is pushed to the browser over WebSockets, so you can watch the lane fill in.
  • Answers are graded by the model and streamed back. Passing a challenge earns XP based on its difficulty, minus a little for each failed attempt.
  • Challenges are stored as vector embeddings, which power semantic content recommendations.

Distributed WebSockets

A horizontally scalable WebSocket system where multiple server instances coordinate through Redis and RabbitMQ for real-time cross-server messaging.

Next project

Infrastructure