EduLab

Bridging the global fluency gap with an AI-humanoid tutor that sees, hears, and researches like a world-class educator.

Comment

Inspiration

Learning IELTS alone is frustrating: you don’t get real feedback, you can’t practice under pressure, and finding good materials takes time. We wanted a single place where students can practice all four IELTS skills with realistic prompts, AI feedback, and optional live tutoring.

What it does

Edulab is an AI-powered IELTS practice platform that helps students train:

  • Speaking : high-pressure examiner-style mock test flow (Part 1 → Part 2 cue card → Part 3), with scoring and actionable fixes.
  • Reading : AI-generated passages + multiple-choice questions with instant scoring.
  • Listening : AI-generated listening scripts converted to AI voice audio , followed by multiple-choice questions and scoring.
  • Writing : Task prompts (Task 1/2) with AI evaluation, band estimate, feedback, and fixes.
  • Live learning : real-time video consultation using Agora for teacher ↔ student practice sessions.
  • Auth + user profiles : Supabase authentication (Google + email/password) and user profile storage.

How we built it

  • We mainly use the Trae IDE for UI Engine. Entire codebase was developed with Trae.
  • Frontend : React + TypeScript (Vite) for a fast, single-page UI.
  • Backend : Node.js + Express REST API.
  • AI + content
  • OpenRouter for question generation, feedback, and evaluation/scoring.
  • Exa for retrieving real-world topics/material inspiration (e.g., recent Speaking Part 2 topics and practice materials).
  • ElevenLabs for text-to-speech, powering Listening audio (and examiner voice support).
  • Realtime video
  • Agora Web RTC for joining/publishing/subscribing to live calls.
  • Server-side token generation with Agora Access Token.
  • Database/Auth
  • Supabase Auth (Google OAuth + email/password).
  • Supabase Postgres for storing user profiles with row-level security policies.

Challenges we ran into

  • Reliable IELTS Part 2 topics : web search results often returned irrelevant pages (like test centers), so we had to filter and extract cue-card-like prompts.
  • Frontend security pitfalls : keeping secrets out of the client build (anything VITE_ is exposed) and moving sensitive keys server-side.
  • OAuth configuration : Google Sign-In required correct authorized origins and redirect URIs, and Supabase provider configuration.
  • Real-time media integration : handling join/leave, publish/unpublish, remote tracks, and UI preview states cleanly.
  • UI/UX polish : making the Speaking experience feel like a real exam with the right pacing, controls, and layout.
  • Running out of credits while testing the web.

Accomplishments that we're proud of

  • Built an end-to-end IELTS 4-skills practice flow with scoring and actionable feedback.
  • Delivered Listening with AI voice audio , which feels closer to real exam conditions.
  • Created a more realistic Speaking mock test with examiner-like pressure and structure .
  • Integrated Supabase authentication and profiles so the app can grow into a real product.
  • Added a working Agora live tutoring foundation for real teaching and learning sessions.

What we learned

  • Mixing web-sourced content with AI generation needs careful filtering to avoid irrelevant or low-quality results.
  • Auth and OAuth are less about code and more about correct platform configuration.
  • Security boundaries are critical: client env variables are not secrets, and keys must be protected server-side.
  • Real-time video is manageable when you keep the “join/publish/subscribe/leave” lifecycle explicit and robust.

What's next for Edulab

  • Persist progress : store attempts, scores, and feedback history per user (dashboards, streaks, goals).
  • Better IELTS-style grading : more detailed band breakdowns (TR/CC/LR/GRA for writing; fluency/lexical/grammar/pronunciation for speaking).
  • Teacher tools : lesson scheduling, notes, shared materials, and post-session feedback templates.
  • More question types : matching headings, fill-in-the-blank, short answers for Reading/Listening.
  • We’re looking to expand beyond IELTS into other linguistics (TOEFL, DET) and professional certifications. We also plan to introduce "Global Peer Mocks," where two students can practice together while the AI acts as a silent moderator and grader. Our ultimate goal is to make world-class language coaching accessible to anyone with a smartphone.

Built With

Share this project:

Updates