Inspiration
India’s classrooms are often crowded, with teachers spending countless hours drafting exam questions and grading assignments. This leaves little time for personalized instruction and timely feedback. We wanted to solve this challenge by building an AI-driven platform that automates these repetitive but essential tasks.
What it does
ShikshaSetu is an agentic AI platform with two specialized agents:
- Question Generation Agent – creates diverse MCQs and subjective questions aligned with curriculum goals.
- Evaluation Agent – grades responses instantly using rubric-based scoring, giving students fair and actionable feedback.
Together, they provide an end-to-end automated assessment workflow that saves teachers hours of work and empowers students with rapid, consistent evaluations.
How we built it
- Designed a Streamlit web app for an intuitive teacher–student interface.
- Integrated large language models for question generation and rubric-based evaluation.
- Built validation layers to ensure MCQs always have four unique options and exactly one correct answer.
- Implemented backend storage with JSON + Pandas for saving attempts, results, and logs.
Challenges we ran into
- Ensuring generated MCQs were valid (no duplicates, no “all/none of the above”).
- Building rubric-based evaluation that remains consistent across varied responses.
- Managing API quotas and handling fallback cases when heavy usage occurs.
Accomplishments that we’re proud of
- A working end-to-end system where teachers can generate, assign, and evaluate exams in just minutes.
- Clean, user-friendly interface accessible from any device.
- Making education technology accessible and scalable for Indian classrooms.
What we learned
We learned how to design agentic AI workflows, build reliable validation pipelines for LLMs, and design interfaces that balance teacher control with AI automation. We also deepened our understanding of how AI can practically support educators at scale.
What's next
- Multilingual support for Indian languages.
- Adaptive learning features that adjust question difficulty based on student performance.
- API integration with other Learning Management Systems for broader adoption.
- Classroom pilots to evaluate real-world impact.
