AI Assistant for teacher

Turn classroom audio into structured insights—summaries, action items, and parent-ready notes in seconds

Comment

Inspiration

What it does

ClassClear — AI Assistant for Teachers in Southeast Asia

🌏 Inspiration

In many Southeast Asian classrooms, teachers often switch between their native language and English while teaching. After each lesson, they spend significant time manually summarizing what was taught, writing action items, and preparing notes for parents.

This repetitive administrative work reduces the time teachers can dedicate to improving teaching quality and engaging with students.

We asked:

What if AI could handle all post-class documentation automatically?

💡 What it does

ClassClear is an AI-powered web application that transforms classroom audio into structured, meaningful outputs:

  • 🎤 Speech-to-text transcription (accent-aware, multilingual)
  • 📝 Automatic lesson summaries & meeting minutes
  • ✅ Action items extraction
  • 😊 Class sentiment analysis
  • 🌐 Translation to English for parent communication

All in just a few seconds after a lesson ends.

🛠️ How we built it

We built ClassClear as a full-stack web application:

Frontend

  • React / Next.js
  • Simple UI for recording or uploading audio
  • Display structured outputs (summary, actions, sentiment)

Backend

  • Node.js + Express
  • Handles audio upload and processing pipeline

AI Integration (VALSEA APIs)

We leveraged multiple VALSEA endpoints to create a complete pipeline:

  1. Transcription API
    Convert classroom audio → text (supports Vietnamese & SEA accents)

  2. Format API
    Transform raw transcript → structured:

    • Summary
    • Key points
    • Action items

  3. Sentiment Analysis API
    Analyze classroom tone → engagement & mood insights

  4. Translate API
    Convert notes → English for parents

Pipeline Flow

Audio → Transcription → Formatting → Sentiment → Translation → UI Output

🚧 Challenges we faced

1. Multilingual & mixed-language input

Teachers often switch between languages mid-sentence.

→ We solved this by:

  • Using robust transcription
  • Designing prompts that preserve context across languages

2. Structuring messy transcripts

Raw transcripts are noisy and unstructured.

→ Solution:

  • Carefully designed prompts for the Format API
  • Enforcing consistent output schema (JSON-like structure)

3. Real-time UX vs processing time

AI processing introduces delay.

→ Solution:

  • Show step-by-step progress (transcribing → summarizing → done)
  • Keep UI responsive and simple

4. Making outputs actually useful

Summaries alone are not enough — teachers need actionable insights.

→ Solution:

  • Focus on action items and parent-ready notes
  • Prioritize clarity over verbosity

📚 What we learned

  • How to design AI pipelines instead of single API calls
  • The importance of prompt engineering for structured outputs
  • Handling multilingual data in real-world scenarios
  • Building AI-first UX that feels fast and useful

🚀 What's next

  • Real-time transcription during class
  • Voice feedback for pronunciation training
  • Integration with LMS platforms
  • Personalized student insights over time

❤️ Impact

ClassClear helps teachers:

  • Save hours of manual work
  • Communicate better with parents
  • Focus more on teaching, less on paperwork

Teach more. Write less. Let AI handle the rest.

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for AI Assistant for teacher

Built With

Share this project:

Updates