Inspiration

Education is evolving, but professors often struggle to gauge real-time student understanding during lectures or identify key learning gaps from homework. Inspired by the idea of using AI to bridge this gap, we set out to build a solution that provides professors with deeper, actionable insights—making classrooms more interactive and helping students succeed.

What it does

ClassPulse is an AI-powered assistant for professors that helps them better understand their students through two key approaches:

  • Real-time Classroom Insights: ClassPulse listens to live lectures and generates questions for the class. It collects student responses and provides immediate insights on how well students are grasping the content.
  • Homework Analysis: ClassPulse grades homework automatically and performs detailed question-level analysis to surface common mistakes and trends. This helps professors pinpoint where students struggle the most and adjust their teaching accordingly.

How we built it

  • Backend: Node.js for building APIs, PostgreSQL for data storage, and OpenAI/Gemini APIs for natural language processing.
  • Frontend: Next for the user interface, offering a dashboard for real-time insights and homework analysis.
  • AI Models: Used OpenAI and Google for grading, text analysis, summarization, and question generation.
  • Architecture: Designed a scalable system, integrating real-time data processing and background jobs for homework insights.

Challenges we ran into

  • Real-Time Performance - one of our biggest technical challenges was optimizing the real-time performance of our system. When a professor creates a learning check in the middle of lecture, we needed to process audio, video, and slides both quickly and reliably.
  • Complexity of AI Workflows - managing multiple AI workflows, from generating contextual questions to grading responses and extracting insights, required careful development to ensure both quality and reliability across different types of lecture content and student responses.

Accomplishments that we're proud of

  • Successfully built an AI system that integrates presentation slides, professor audio, and the video recording to generate questions and provide insights during lecture in real-time.
  • Developed advanced NLP pipeline that grades questions, classifies errors, and identifies gaps in student knowledge, allowing professors to make data-driven decisions.
  • Created a full platform experience, including a user-friendly interface that makes complex insights easy to understand and visualize

What we learned

  • Multi-modal data processing - techniques for summarizing large amounts of diverse educational data into concise, actionable insights
  • AI system architecture - designing scalable systems that orchestrate multiple AI models while maintaining reliability
  • Full-Stack Integration - bridging modern frontend frameworks and visualizations with AI-powered backends to create a seamless, responsive classroom platform

What's next for ClassPulse

  • Enhanced Student Insights: Expand analysis to include long-term trends in student performance across multiple assignments and lectures.
  • Student Engagement Tools: Introduce interactive features for students, such as personalized feedback and progress tracking.
  • Broader AI Model Support: Integrate new AI models like Deepseek and Perplexity for even more accurate question generation and grading.

Built With

Share this project:

Updates