AutoEval

• 

  landing page

• 

  Dashboard

• 

  MCQs

• 

  Create Test

• 

  Upload student submittions

Inspiration

As a high school student, I saw how inefficient and time-consuming traditional assessment methods can be. I wanted to create a system that automates the entire evaluation workflow. from generating question papers to personalized student feedback, freeing up teachers to focus more on instruction.

What it does

AutoEval automates question paper generation, scans and grades answer sheets, analyzes submissions, and generates personalized notes for students based on their mistakes. It uses the Perplexity AI API for intelligent feedback generation and analysis.

How we built it

The platform was built using Django for the backend, Tailwind CSS for the frontend, PostgreSQL for the database, and Docker for containerization. For scanning and grading, I integrated the open-source OMRChecker tool. I solved its rigid input structure by creating dynamic blueprint folders that simulate the expected format, then deleting them post-processing to optimize storage.

Challenges we ran into

OMRChecker required a specific folder layout with a template.json and scanned images. Automating this while keeping the system stateless and storage-efficient was a major hurdle. I resolved this by programmatically generating temporary folders during each submission and cleaning them up immediately after use.

Accomplishments that we're proud of

• End-to-end automation of the evaluation process

• Integration of AI for meaningful, personalized feedback

• Solving storage and I/O bottlenecks with dynamic folder management

• Building a scalable, containerized system as a high school student

What we learned

• Practical integration of AI APIs into real-world applications

• Managing file I/O and storage efficiently in a web app

• Dockerization

• How to overcome rigid input formats of legacy ML tools

What's next for AutoEval

• Add support for different question formats beyond OMR (e.g., typed, written)
• Integrate machine learning and AI-powered analytics for deeper insights
• Conduct studies on student submissions to identify learning patterns and improve pedagogy
• Add handwriting recognition for descriptive answers
• Public beta release for schools to pilot

Built With

• django
• omrchecker
• postegresql
• python
• tailwindcss