Asisten Dosen App

A cloud platform that helps university lecturers and educators automate academic administration, lesson planning, and research tasks that powered by AI.

Comment

Inspiration

Repetitive administrative tasks, course assessments, and lesson planning consume time of the lecturer that should be spent on more productive. The goal was to create an agentic AI platform to automate the academic and research assistant.

What it does

This App provides a solution in academic and research as assistant for lecturer.

  • *RPS Generator * Automatically creating semester lesson plans conforming to to the rule.
  • HOTS Question Bank:Evaluation questions based on Bloom's Taxonomy (C1–C6) generation
  • Content Generation: Creating PPT slide outlines and comprehensive lecture notes
  • Auto-Grader: Automatically grades student essay answers based on a defined rubric
  • Research Assistant: Helps identify research trends, gaps, and drafts proposals.

How we built it

We utilized a modern, serverless architecture:

  • Frontend: A Single-Page Application (SPA) built with Vanilla JS and Navigo Router, featuring a Glassmorphism design, deployed on Firebase Hosting.
  • Backend: Built with FastAPI (Python 3.11) and deployed as a serverless container on Google Cloud Run
  • AI Agent Layer: Powered by Gemini 3.5 Flash utilizing the Google GenAI SDK, featuring prompt chaining and a RAG Context Builder.
  • Cloud Infrastructure: We leveraged Google Cloud Firestore for our NoSQL database, Cloud Storage for secure file handling, and Cloud Pub/Sub for managing asynchronous background tasks.

Challenges we ran into

One of the main challenges was engineering the prompt chaining to ensure the AI's output strictly adhered to the complex formatting of the outcome that comply the rule and standard. Additionally, orchestrating long-running AI generation tasks required us to implement a robust asynchronous queue system using Google Cloud Pub/Sub and FastAPI workers to prevent timeouts.

Accomplishments that we're proud of

We are incredibly proud of building a fully serverless, highly scalable architecture. We successfully developed a responsive Single-Page Application using pure Vanilla JS[cite: 1] without relying on heavy frontend frameworks, while maintaining a modern Glassmorphism UI[cite: 1]. We also implemented a secure RBAC (Role-Based Access Control) system and a GDPR-compliant data architecture[cite: 1].

What we learned

We deepened our understanding of agentic AI workflows, specifically how to use the Google GenAI SDK to chain prompts and build RAG contexts effectively. We also learned how to seamlessly integrate Google Cloud Pub/Sub push triggers with Cloud Run for reliable background processing

What's next for Asisten Dosen App

Our next step is to build direct REST API and webhook integrations so the Auto-Grader and Assignment Triage can connect seamlessly with popular Learning Management Systems (LMS). We also plan to expand the Research Assistant's capabilities by enabling it to autonomously pull and synthesize literature from external academic databases.

Built With

Share this project:

Updates