For most of us, learning English has been a source of anxiety, not joy. We spent 12 years in school memorizing grammar rules, dissecting sentences like biology specimens, and stressing over exams.
Yet, when we try to speak, we freeze. We translate in our heads. We fear making mistakes. The educational system treated language like a subject to be studied, not a skill to be lived. It asked too much of your logic, and too little of your intuition.
Think back to how you learned your first language (Amharic, Oromiffa, Tigrinya).
- Did you study a textbook at age two? No.
- Did you take a grammar exam at age three? No.
- You simply lived.
You listened to stories. You watched people. You guessed meaning from context. You made mistakes, and nobody graded you "F". You absorbed the patterns unconsciously.
Real English is a digital environment designed to recreate that natural, child-like state of acquisition. This backend acts as the invisible orchestrator—powering interactive narratives, serving addictive cultural video feeds, and tracking organic growth—without ever forcing the user to take a grammar test.
We replaced traditional tests with Playful Narratives.
- The Engine: Manages complex, graph-like JSON structures (
StoryTrail,StorySegment,SingleChoiceChallenge) dynamically generated using the Gemini AI API. - The Tech: Supports branched storytelling, choice validation with contextual feedback, and tracks user progression. It uses the Gemini API for Text-to-Speech (TTS) audio generation, Pollinations AI for scene image generation, and Ethio Telecom (Tele) Cloud OBS (S3-compatible) to store and serve these generated audio and image assets directly to the mobile client.
We replaced lectures with Addiction.
- The Engine: Powers the TikTok-style vertical video feed by curating learning shorts (harvested from YouTube). It categorizes clips (funny, real-life, motivation, culture) and tailors the experience by difficulty.
- The Tech: Tracks user watch history (
UserShortHistory), manages saved videos for later spaced repetition, and delivers personalized infinite-scroll feeds with smart tag filtering.
We replaced grades with Nature.
- The Engine: An organic progression system centered on streaks, active days, and leveling up ("Tree Stage"). Includes secure, frictionless onboarding.
- The Tech: Implements Google Sign-In, OTP integrations, JWT secure sessions, and profile tracking (Avatar logic, streak calculations). Uses Nodemailer and AWS integration for account recoveries securely.
The backend follows Clean Architecture combined with a Feature-First directory structure. This ensures absolute separation of concerns, scalability, and high testability across all moving parts.
Every feature isolates:
- Domain Layer: Business Entities (e.g.,
ImmersionShort,User) and pure logic. No external dependencies. - Use Cases (Application): Orchestrates the data flow (e.g.,
GetPersonalizedFeed,GetNextStoryTrail). - Interface Adapters: Express routes, controllers, and middlewares.
- Infrastructure: PostgreSQL repositories (
Prisma), AI clients (Gemini API,Pollinations AI), email services (Nodemailer), and file storage (Ethio Telecom Cloud OBS).
src/
├── features/
│ ├── authentication/ # Sign Up, Sign In (OTP/Google Auth)
│ ├── daily_immersion/ # Video Feed Logic, YouTube Harvest, History
│ ├── profile/ # User Growth & Garden (Streaks, Levels)
│ └── story_trails/ # Interactive Story Engine, Audio delivery
├── di_container.ts # Dependency Injection Container
└── server.ts # Entry Point, Middleware, Routes