Shayak.
Google Classroom, rebuilt with real-time CNN engagement detection
Real-time engagement detection in 6 classes via TensorFlow CNN
Auto-attendance from join/leave + engagement signal
Adaptive quizzes, daily practice problems & refresher AI
120+ Framer-Motion scroll animations with spring physics
A full educational platform that replaces Google Classroom with a tighter, AI-native experience. Frontend in Next.js 14 (Tailwind, shadcn/ui, Framer Motion with 120+ scroll animations, Zustand, TanStack Query). Backend in Node/Express with MongoDB Atlas, JWT auth, Socket.IO for live features, AWS S3 for media, node-cron for scheduled jobs. The differentiator is a TensorFlow/Keras CNN that runs in a separate Flask microservice and classifies student engagement live during video classes into 6 buckets — Actively Looking, Confused, Talking to Peers, Distracted, Bored, Drowsy — feeding both attendance and per-student engagement analytics back to the teacher dashboard.
