Inspiration
The inspiration came from noticing how difficult it is for students to find reliable, CBSE-accurate sample papers that actually feel like real school examinations. Existing resources were either outdated, not aligned with the latest patterns, or lacked diagrams, graphs, and structured marking schemes. As someone who actively prepares for Class 11 and 12 exams, I realized that generating high-quality question papers on demand could save students hours of searching and give them a more realistic practice experience. That idea grew into the foundation of this project.
What it does
The project generates complete Class 11 and 12 CBSE-style sample question papers using AI. It supports all major subjects including Physics, Chemistry, Mathematics, Computer Science, and English. It follows official CBSE patterns, section structures, marks distribution, and difficulty levels. The system also allows optional inclusion of diagrams, graphs, and image-based questions. Users can generate papers, save them, download them, and review them later. The goal is to make practice efficient, accurate, and instantly accessible.
How we built it
The backend uses the Gemini Flash model through an API. A structured, high-detail 9000 token prompt defines the exact CBSE format, ensuring section layout, marks, and question types match real exam papers. The frontend is built as a dashboard with separate views for profile, paper generation, and saved papers. The system fetches user inputs such as subject, class, chapters, difficulty settings, and diagram requirements. These parameters are passed to the AI, which outputs a formatted sample paper. Additional scripts handle saving, displaying, and exporting generated papers.
Challenges we ran into
The largest challenge was quality control. AI models often generate content that drifts away from the exact CBSE exam style, especially in English and long questions. Ensuring structure accuracy, realistic language, and diagram support required multiple revisions to the prompt and logic. Another challenge was API limits, which made the system difficult to scale. Adding image-based questions also required external solutions that could pair diagrams from the web with AI-generated prompts. Maintaining consistency across subjects and chapters demanded precise prompt engineering.
Accomplishments that we're proud of
We successfully created a fully functional AI system that can generate exam-ready CBSE sample papers in seconds. It handles PCM, CS, and English with high accuracy and produces questions that match the tone and complexity of real board-level exams. The dashboard allows users to store and organize their generated papers. The system also managed to include diagram based questions with structured formatting. Building all this as a student project with limited resources is something we are proud of.
What we learned
We learned how important prompt design is when working with large language models. Even small changes in instructions can significantly impact the quality of generated exam papers. We also learned about handling API constraints, optimizing requests, and managing user workflow in an educational tool. The project taught us how to balance accuracy with scalability, and how to design an interface that supports real study behavior rather than just showcasing AI capabilities.
What's next for Hackathon
Next steps include improving diagram generation using a hybrid approach of web search and synthetic image creation, adding support for more subjects, and building teacher-mode features where educators can upload their own questions. We also plan to introduce a subscription model, integrate UPI or Paytm for Indian users, and optimize AI output so the papers feel indistinguishable from real CBSE sample sets. Long term, this could evolve into a full platform for AI generated assessments, quizzes, and study tools.
Log in or sign up for Devpost to join the conversation.