ClassSense AI

1. Landing page of the app.
2. Chat Upload: bar chart for understanding, clarification urgency, and teaching suggestions
3. Clustered Doubts(similar doubts asked by multiple students) and Uncommon doubts.

✨ Inspiration

Online classes and bootcamps have become a part of everyday learning. However, in large online classes, teachers often miss student doubts. Chats move too fast, the same question is asked in different ways, and polls interrupt the natural flow of teaching.

While attending live classes myself, I noticed that many genuine doubts were never addressed — not because the teacher didn’t care, but because the cognitive load of managing chat was too high. This inspired me to build ClassSense AI, a tool focused on simplifying everyday teaching workflows.

🧠 What I Learned

Building ClassSense AI helped me learn:

How large language models like Google Gemini understand informal, real-world language
Why scope control is critical during hackathons
How AI can be used to reduce cognitive load, not increase it
How to handle real-world issues like API quotas, SDK deprecations, and tooling mismatches

I learned that a small, well-defined solution is often more impactful than a feature-heavy one.

🛠️ How I Built the Project

ClassSense AI is a lightweight, end-to-end AI system:

A teacher pastes or uploads classroom chat text
The chat is analyzed using Google Gemini
The AI estimates student understanding, clusters similar doubts, and highlights unique questions
Results are displayed in a clean dashboard built with Streamlit

Instead of keyword matching, the system relies on semantic reasoning, allowing it to understand differently worded doubts with the same meaning.

model = genai.GenerativeModel("models/gemini-flash-lite-latest")
response = model.generate_content(prompt)

This approach makes the tool fast, flexible, and suitable for everyday classroom use.

🚧 Challenges I Faced

Some key challenges during development were:

Managing model availability and free-tier quota limits
Handling SDK deprecation warnings and version mismatches
Ensuring AI output was structured and safe to display Solving these challenges required iterative debugging and fallback strategies, similar to real-world engineering problems.

🚀 What’s Next

In the future, ClassSense AI can integrate directly with platforms like Zoom or Google Meet to provide real-time classroom insights. This would enable live understanding tracking and adaptive teaching support during ongoing lectures.

🏁 Conclusion

ClassSense AI shows how AI for Everyday Life (TRACK 1) can simplify daily workflows. By reducing noise, surfacing clarity, and saving time, it helps teachers focus on what truly matters — effective teaching and learning.