RetinaRank

RetinaRank - AI Quiz Generator & Feedback
Project Flow - Quiz Generation to Result Report

Inspiration

Current quiz systems lack a critical feature: they can't verify if students are honestly attempting assessments. This gap inspired us to build a solution that brings integrity to online testing.

What it does

RetinaRank combines computer vision with AI to create an intelligent quiz platform. It uses eye-tracking technology to monitor user focus and engagement in real-time, ensuring academic honesty. The platform leverages Gemini to automatically generate quizzes and evaluate responses, creating a seamless end-to-end assessment experience.

How we built it

We utilized a powerful tech stack including:

AI Tools: Cursor, Claude, GitHub Copilot, and Lovable for accelerated development
AI Engine: Gemini 3 API for quiz generation and evaluation
Computer Vision: Custom eye-tracking implementation for focus detection

Challenges we ran into

Building a document-to-markdown converter from scratch
Integrating with the Gemini API and handling its responses
Coordinating code integration across the team
Managing deployment complexities

Accomplishments that we're proud of

We created a platform with genuine potential for global impact—one that addresses a real problem in online education and makes remote assessments more trustworthy.

What we learned

AI models process markdown natively, leading us to develop our document-to-markdown converter
How to effectively work with the Gemini API
Practical applications of the Pydantic library in Python for data validation
The concept and power of "vibe coding"—rapid AI-assisted development

What's next for RetinaRank

We envision adding a teacher dashboard where educators can create custom quizzes while students take them remotely. The system would provide instructors with real-time engagement analytics and cheating detection reports, making remote proctoring both scalable and effective.

Built With

Submitted to

Gemini 3 Hackathon

Created by

I worked as the project lead and as AI Engineer. In Simple words: i worked at the backend handling gemini API calls, taking the .md user notes and generating practical quiz questions in JSON format, ready to be attempted at the frontend.

Soyyam katyara
I worked on back end activity that is covert file to markdown that is perfect for AI to recognize and I used pdf libraries in python and I learned a lot from them

Varoon Kumar
I worked on building the AI-powered quiz generation feature using the Gemini 3 API. My role involved designing prompts, integrating the API, and generating dynamic, context-aware quiz questions in real time. I focused on ensuring question quality, relevance, and difficulty balance while keeping response times fast. This helped automate quiz creation and made the learning experience more engaging and personalized. Through this, I gained hands-on experience with LLM integration, prompt engineering, and handling AI-generated content effectively.

Gotam Talreja
Contributed to the computer vision module by researching and evaluating eye and retina tracking techniques. Explored multiple models and approaches for detecting ocular features and tracking eye movement, supporting the selection of suitable methods for the project.

Ali Quitb Sufi
I worked on a computer vision–based system to detect eye focus, aimed at understanding where and how a person’s attention is directed. The project involved using camera input to capture facial and eye-region data, followed by preprocessing steps such as face detection, eye localization, and noise reduction to improve accuracy. I applied eye-tracking and gaze-estimation techniques to analyze eye movement patterns and determine focus direction in real time. The system leveraged computer vision algorithms and machine learning models to classify focused versus unfocused states, making it useful for applications like attentive monitoring, human–computer interaction, and behavioral analysis. Overall, this work strengthened my understanding of image processing, feature extraction, and practical deployment of computer vision solutions.

Haseeb Shah
I worked on backend activity that converts files (DOCX, PPT, PPTX) into Markdown format,making it easier for AI to recognize and process,I used Python libraries to implement this.
Through this work, I learned a lot about file parsing,text extraction, and structured data conversion

Yash Raj
I managed the Evaluation Phase, creating the API flow that compares user answers against the quiz key using Gemini 3 Pro. I designed the AI prompts for fair grading and structured the JSON results to combine academic scores with "focus level" data for a complete performance report.

Vishal Jeevan
Abdul Ghaffar
vinesh batheja
Shoaib Ahmed