Class Sense

Main page
Library
Engagement risks
Micro-Assessments
Parent Summary
PDF export

Inspiration

Teachers spend hours preparing lessons yet still struggle to predict where students will get confused. We wanted an AI assistant that analyzes lesson materials before class and highlights where learning might break down, saving time and improving clarity.

What it does

ClassSense lets teachers paste or upload lesson content and instantly generates:

Engagement/confusion hotspots with risk levels (High/Medium/Low)
Section-level risk analysis with specific reasons and solutions
Micro-assessments (multiple choice, short answer, discussion questions)
Parent-friendly summaries with email-ready templates
Differentiation strategies for ESL, struggling, and advanced learners
Lesson clarity score (0-100) with visual progress indicator
Lesson library to save and revisit past analyses
Professional PDF exports with color-coded risk badges

All delivered in a clean, teacher-friendly dashboard with three focused tabs.

How we built it

Frontend: React 19 + TypeScript + Vite for blazing-fast development
Styling: Tailwind CSS with Lucide icons for a modern, clean interface
Visualization: Recharts for interactive hotspot bar charts
AI Engine: Google Gemini 2.5 Flash with Structured Output API for guaranteed JSON consistency
PDF Generation: jsPDF for client-side professional reports
Storage: Browser localStorage for offline-capable lesson history

The app is a lightweight, fully client-side application with zero backend infrastructure—just one API call to Gemini for analysis.

Challenges we ran into

Structuring AI prompts to generate actionable, evidence-based insights (not generic suggestions)
Designing a schema that works with Gemini's Structured Output API for consistent JSON responses
Creating a clarity score algorithm that meaningfully represents lesson quality
Building a multi-page PDF generator with proper page breaks, styling, and color-coded badges
Making the analysis fast within hackathon constraints while maintaining quality
Balancing feature richness with a clean, non-overwhelming teacher UX

Accomplishments that we're proud of

Built a production-ready MVP with polished UX and real classroom value
Created multi-dimensional analysis (cognitive load, pacing, prerequisite detection)—not just content summarization
Achieved zero backend complexity—entire app runs client-side with localStorage
Implemented Gemini 2.5 Flash's Structured Output API for bulletproof JSON schemas
Designed parent communication bridge that turns technical lessons into friendly email summaries
Delivered professional PDF exports with custom styling, risk badges, and multi-page support
Built a lesson library system for teachers to revisit and compare past analyses

What we learned

How teachers think about cognitive load and where students actually struggle (not where we expect)
The power of Gemini's Structured Output API for eliminating AI hallucination risks
How to design evidence-based differentiation strategies grounded in actual lesson content
The importance of small UX touches (skeleton loading, demo content, copy-to-clipboard) for teacher adoption
That client-side architecture can deliver serious value without server complexity
How clarity scoring can make abstract lesson quality tangible and actionable