Langgit

Bite-sized language learning lessons, delivered by a virtual teacher & personalised based on your reddit interests.

Comment

Inspiration

With recent advances in AI, interactive virtual teachers are now within reach, This unlocks the potential for highly interactive, personalised learning and for very low cost.

What It Does

Langgit delivers gamified mini-lessons led by realistic AI teachers, tailored to each use. Each lesson is designed to be quick, fun, and effective, ensuring a dynamic and personalized learning experience.

Student can complete mini lessson based on their interest and upload h.w on PDF based on materials that want to study

  • Bite-Sized Lessons: Each "Langgit" can be completed in just 5 minutes, providing an accessible way to learn that fits into busy schedules.
  • Daily Langgits: Unique AI-generated content is delivered daily, encouraging users to return regularly. By incorporating features like a streak leaderboard, Langgit rewards users for consistency and dedication.
  • Community Focus: Langgit targets millions of language learners across Reddit’s language learning communities, offering tailored experiences for dozens of languages.

How We Built It

  • Frontend: Built with Vite/Vue.js
  • Backend: Node.js & Python
  • Custom AI Backend: Includes Voice TTS (Text-to-Speech) and Lip Sync workflows.

AWS Services

  • **S3 buckets for media storage
  • Lambda for Data prep & Post processing - Bedrock for LLM course and content creation , and generative images for quizzes and course visuals -** RDS for DB backend -** Amplify hosting /frontend -** Polly for TTS

Challenges We Encountered

  • Efficient AI Inference: AI inference cost vs speed.
    Inference was either too slow or cost was too high so we had opted to move away from realtime generation for most tasks

Accomplishments We're Proud Of

Originally, the plan was to generate personalized images and audio in near real-time. However, due to inference times, and costs, We pivoted to generating personalised mini-classes for the following day. This approach not only encourages daily engagement but also significantly reduces inference costs by queuing resource-intensive tasks like lip-syncing.

In the future, as the library of AI-generated content grows, we can reuse and combine existing materials across the user base. This will further minimize inference time and costs while reintroducing instant access to personalized content."

What's Next for Langgit

  • P2P Practice & Real-Time Collaborative Games: Introducing features for peer-to-peer language practice and role-playing exercises with avatar overlays.
  • Premium Version: Launching a subscription-based model.
  • Community Partnerships: Collaborating with language learning communitie to expand our reach and engagement.

Built With

Share this project:

Updates