About the Project
RubriqAI is an AI-powered feedback system designed to help students understand why they received a certain grade and how they can improve, not just what score they earned.
The inspiration for RubriqAI came from a simple but persistent problem in education: feedback is often slow, inconsistent, and unclear. Many students receive a grade with minimal explanation, leaving them unsure of what they did wrong or how to improve next time. At the same time, teachers face increasing workloads, larger class sizes, and limited time to provide personalized feedback for every student.
RubriqAI aims to bridge this gap by making high-quality, rubric-aligned feedback scalable.
What Inspired Me
As a student, I’ve often received feedback that felt generic or vague, comments like “add more detail” or “needs improvement” without specific guidance. I realized that the issue wasn’t that teachers didn’t care, but that providing detailed, personalized feedback for every assignment simply doesn’t scale.
I wanted to explore whether AI could be used responsibly, not to replace teachers, but to support learning by giving students clearer, more actionable feedback tied directly to the rubric.
How the Project Works
RubriqAI analyzes three inputs:
The assignment rubric
The assignment instructions
The student’s submission
These inputs are uploaded as PDFs. The server then sends them to Gemini AI through a carefully engineered prompt that treats the rubric as the single source of truth.
The AI generates a structured feedback response, which the server:
Parses from JSON
Extracts key elements such as grades and rubric coverage
Beautifies and displays as a clear feedback report for the student
This process requires no training data, no manual setup, and works across subjects and grade levels.
How I Built It
RubriqAI was built as a full-stack web application using:
Frontend: HTML5, CSS3, Bootstrap, JavaScript
Backend: Python, Flask
AI Integration: Gemini API
Document Processing: PyPDF2
The UI was designed with a glassmorphism aesthetic, offering both Light and Dark Mode while maintaining accessibility standards such as WCAG contrast, keyboard navigation, and large tappable targets.
Challenges I Faced
One of the main challenges during development was connecting the server to the Gemini API and reliably parsing the structured JSON responses it generated. Because RubriqAI depends on consistent output, i had to ensure that the AI’s response format could be safely interpreted, processed, and displayed without breaking the user experience.
Another challenge was implementing Light and Dark Mode functionality while maintaining visual consistency and accessibility. This required careful design decisions to ensure readability, contrast, and usability across themes without impacting performance.
Finally, ensuring consistent rubric-aligned feedback was a critical challenge. Since rubrics can vary widely in structure and clarity, I had to design prompts and parsing logic that kept the rubric as the single source of truth while still producing meaningful, personalized feedback.
Overcoming these challenges helped transform RubriqAI from a basic prototype into a stable, scalable system.
What I Learned
Through this project, I learned how to:
Integrate large language models into real-world applications
Design systems that prioritize learning over grading
Build accessible, polished user interfaces
Engineer prompts that produce consistent, structured AI output
Think critically about ethical AI use in education
Final Reflection
RubriqAI is not about automating grading it’s about guiding improvement. By aligning feedback directly with rubrics and explaining both strengths and weaknesses clearly, the project demonstrates how AI can be used to enhance learning outcomes while keeping teachers at the center of the process.
This project represents my interest in building technology that is not only proficient, but also meaningful and impactful in real-world settings.
Log in or sign up for Devpost to join the conversation.