km.ai - ketua murid AI

Inspiration

Teachers often struggle to manage, grade, and monitor the progress of large classes while also handling administrative tasks such as scoring exams and compiling student reports. KM.ai was created to help reduce this workload through automation and intelligent assistance.

What it does

KM.ai helps teachers generate lesson materials, create randomized exam sets, and evaluate student answers automatically. It also provides a clear overview of class and individual progress, allowing teachers to focus on personalized teaching instead of manual administrative work.

How we built it

The system is fully serverless, built on Google Cloud Run using:

• Cloud SQL for structured data storage
  
• Pub/Sub for asynchronous event handling
  
• Google Generative AI for content generation and grading
  
• React + Shadcn + Vite for the teacher dashboard interface
  

Challenges we ran into

• Managing Pub/Sub message flow and ensuring consistent event delivery
  
• Designing effective AI prompts that produce reliable results for varied question types
  

Accomplishments that we're proud of

• A fully serverless, event-driven pipeline that scales automatically
  
• Cost-efficient batch AI processing
  
• Clean modular architecture deployable with minimal infrastructure management
  

What we learned

• Building event-based systems on Cloud Run
  
• Integrating Google Generative AI effectively in real-world workflows
  
• Structuring multi-service apps for scalability and low latency
  

What's next for KM.ai – Ketua Murid AI

Future plans include adding agentic consultation features that allow teachers to receive real-time insights and teaching recommendations based on class performance trends.

Built With

• cloudrun
• google-generative-ai
• google-studio
• nestjs
• node.js
• postgresql
• react
• typescript