Case Study — AI Automation
Digital Classroom Insights: On-Device Emotion AI for Classrooms
An AI-powered emotion detection system built for real classroom environments, with a strong focus on privacy through secure, entirely client-side processing.
Overview
Overview
Digital Classroom Insights is an AI-powered emotion detection system built for real classroom environments, live at dcistudents.app.
Problem
Problem
Instructors need real-time visibility into how engaged students are — but most emotion-detection systems either send video to a server, which raises real privacy concerns in a classroom, or rely on ML models too large to run in real time on the edge.
Solution
Solution
BM Automate built a system with secure, client-side emotion detection — no video ever leaves the device — and optimized the underlying ML model itself, reducing it from 100MB down to 3.3MB for edge deployment while maintaining 70% accuracy.
Key Features
Key features
- Real-time emotion detection running entirely client-side via TensorFlow.js
- Live video capture via WebRTC, processed on-device — video is never uploaded
- Edge-optimized ML model: reduced from 100MB to 3.3MB while maintaining 70% accuracy
- Node.js and PostgreSQL backend for aggregated engagement data, not raw video
- Clerk-based authentication for instructor accounts
Tech Stack
Tech stack
Result
Result
A roughly 30x smaller ML model (100MB → 3.3MB) suitable for real-time edge deployment, while maintaining 70% accuracy — with a privacy-first architecture where video never leaves the student's device.
Visit dcistudents.app→Need something similarly privacy-conscious?
Tell us what you're building and we'll tell you exactly what it takes to build it right.
Book a Call