Building in Public · Early Access Coming Soon

AI that sees, hears &
understands every classroom

Two AI agents — ARIA generates the questions shy students won't ask, ATLAS observes what teachers physically can't. Camera + microphone powered classroom intelligence.

Camera + Mic Powered Privacy-First Architecture Real-time AI Analysis

The Problem

Teachers Can't See What They Can't See

75%

of students are too anxious to ask questions in class

Education Week Research, 2023

~1.5 min

of individual attention per student in a 30-student, 45-minute class

Classroom time arithmetic

3-6 weeks

average delay before teachers identify struggling students

Fuchs & Fuchs, 2004

See It In Action

Watch AI Agents Work in a Live Classroom

Press play to see how ARIA generates questions and ATLAS monitors engagement — in real-time during an actual Biology lesson. No scripted demo — this is what teachers see.

Live Classroom Session

Biology — Class 10 — Cell Transport Mechanisms

ARIA Active
ATLAS Monitoring
Engagement: 85%
0:003:00

ARIA is listening...

Questions will appear as the lesson progresses

The AI Agents

Two Agents, One Smarter Classroom

ARIA

Artificial Reasoning & Inquiry Agent

ARIA listens to every lecture via classroom microphones, identifies knowledge gaps the moment they occur, and generates the questions students are too shy to ask — ranked by Bloom's Taxonomy for pedagogical depth.

Powered by: Classroom mic array → real-time speech-to-text → LLM analysis

Real-time Transcription
Question Generation
Bloom's Taxonomy Tags
Comprehension Scoring
Knowledge Gap Detection
Speaker Diarization

ATLAS

AI Teaching & Learning Analytics System

ATLAS uses classroom cameras to observe what teachers physically can't — individual student attention, body language, hand raises, and participation patterns. It provides real-time coaching nudges so teachers can adapt on the fly.

Powered by: Classroom cameras → computer vision → engagement ML models

Attention Tracking
Body Language Analysis
Hand Raise Detection
Live Teacher Coaching
Participation Heatmaps
Behavior Pattern Alerts

What You'll Need

The hushClass Kit

Real classroom intelligence requires real sensors. hushClass combines affordable hardware with powerful AI to give teachers genuine visibility into their classrooms.

Classroom Camera

A wide-angle camera positioned to observe the full classroom. Can be a dedicated hushClass camera or your existing CCTV.

  • Wide-angle lens (120°+ FOV)
  • 720p minimum resolution
  • Student attention & posture tracking
  • Hand raise detection
  • Privacy: processed on-device, no raw video stored

Microphone Array

A directional mic array that captures classroom audio for real-time transcription and speaker identification.

  • Ceiling or desk-mountable
  • Noise cancellation for classroom acoustics
  • Speaker separation (teacher vs student)
  • Real-time audio streaming
  • Works with standard classroom layouts

hushClass Hub (Optional)

An edge compute device that processes camera and audio feeds locally — keeping raw data private and reducing cloud dependency.

  • Connects to camera + mic
  • On-device video processing
  • Only metadata sent to cloud
  • Works with limited internet
  • Plug-and-play setup

Software-only mode: Schools with existing CCTV infrastructure can integrate directly. Teachers can also start with audio-only mode using a laptop microphone — camera features are additive, not required.

Student Identification

How hushClass Knows Who Is Who

For ATLAS to track individual engagement, it needs to identify each student. We offer three approaches — schools choose what fits their comfort level.

Most Accurate

Face Enrollment

During onboarding, each student's photo is captured via the hushClass camera. Face embeddings (not photos) are stored securely. The camera identifies students automatically in future sessions.

Privacy-First

Seating Map

Teacher assigns fixed seats in the hushClass app. The camera identifies students by their position — no biometric data needed. Works well with consistent seating arrangements.

Simplest

Student Badges

Each student gets a small desk badge with a QR code or NFC tag. The camera reads badges to identify students. Zero biometric data, easy to set up.

Privacy by Design

Parental consent is required for all enrollment methods. Face embeddings are mathematical vectors — not photos — and are encrypted at rest. Schools retain full control over data retention and can delete all student data at any time. All processing follows privacy-first principles.

How It Works

From Setup to First Insights

Install the hardware, enroll students, and let AI do the rest. Teachers focus on teaching — hushClass handles the observation.

1

Install hushClass Kit

Mount the camera and mic in your classroom. Connect to the hushClass Hub or directly to your school's network.

2

Enroll Students

Import your class roster. Choose face enrollment, seating map, or badge-based identification for each student.

3

Start a Session

Teacher clicks 'Start Session'. ARIA begins listening, ATLAS starts observing. Both agents work silently in the background.

4

Review Insights

After class, review the AI-generated debrief: engagement data, generated questions, comprehension gaps, and coaching suggestions.

Capabilities

What hushClass Actually Measures

With camera + microphone data, here's what becomes possible — no guesswork, just real classroom intelligence.

Via Camera (ATLAS)

Individual Attention Tracking

Head orientation and gaze direction per student — who's looking at the teacher vs their phone.

Body Language Signals

Posture analysis: leaning forward (engaged), slouching (disengaged), note-taking detection.

Hand Raise Detection

Automatic detection when students raise hands. Tracks who gets called on and who doesn't.

Participation Patterns

Which students interact during group activities, who's isolated, who dominates discussions.

Teacher Movement Heatmap

How the teacher moves around the room — which areas get more attention.

Via Microphone (ARIA)

Real-time Transcription

Live speech-to-text of the entire lecture with timestamp alignment.

AI Question Generation

ARIA identifies concepts students likely struggle with and generates clarifying questions ranked by Bloom's Taxonomy.

Speaker Identification

Distinguishes teacher voice from student voices. Tracks which students speak up.

Engagement Proxy via Audio

Classroom noise patterns, silence duration, and energy levels as indicators of engagement.

Concept Coverage Analysis

Maps lecture content against curriculum to identify what was taught vs what was missed.

For Every Stakeholder

From Individual Classrooms to Entire Districts

hushClass is designed for the realities of education — large class sizes, diverse learners, limited observation bandwidth. One system that serves every stakeholder differently.

For Teachers

Real-time coaching nudges during class. Post-session debrief cards with AI-generated questions, engagement data, and teaching suggestions.

For Principals

School-wide engagement analytics, teacher development insights, and automated weekly reports — without sitting in every classroom.

For Students

AI-generated questions that voice doubts they're afraid to ask. Better-paced lessons. More interactive classrooms.

For Counsellors

Early warning alerts for students showing sustained disengagement or behavioral pattern changes — weeks before grades drop.

Audio-only

Minimum Mode

Camera + Mic

Full Intelligence

On-premise

Edge Processing

Zero PII

Raw Video Storage

Architecture

Built to Integrate, Not Replace

Existing CCTV Integration

Already have cameras? hushClass can connect to standard IP cameras and RTSP feeds. No new hardware needed for schools with existing infrastructure.

LMS Connectivity

Import class rosters and curriculum data from Google Classroom, Canvas, or any LMS with API access.

Analytics Dashboard

Deep insights on engagement, comprehension trends, and teaching effectiveness. Session-by-session and longitudinal views.

Student Information Systems

Sync with your school's SIS for student enrollment. Auto-import class lists and student IDs.

Parent Communication

Automated engagement summaries and comprehension reports shared with parents at configurable intervals.

Curriculum Mapping

Upload your syllabus. ARIA maps lecture content against planned topics to show coverage gaps and teaching pace.

Pricing

Simple, School-Friendly Pricing

We're co-building pricing with our pilot schools. Early partners get locked-in founder rates. No per-student fees, no surprise overage charges.

Starter

Freeduring alpha
  • 1 classroom
  • Audio-only (no camera required)
  • ARIA question generation
  • Post-session debrief
  • Email support
Start Free
MOST POPULAR

School

Custompricing

Tailored to your school size and needs

  • Up to 20 classrooms
  • Camera + mic intelligence
  • ARIA + ATLAS agents
  • Real-time teacher coaching
  • Principal dashboard
  • Counsellor alerts
  • Priority support
Contact for Pilot Pricing

District / Chain

Custompricing
  • Unlimited schools
  • Dedicated infrastructure
  • On-premise deployment option
  • Custom integrations (SIS, LMS)
  • SLA guarantee
  • Dedicated account manager
  • Training & onboarding
Talk to Us

Prices in INR. Pilot partners lock in founding rates permanently. No per-device fees. No lock-in during alpha.

Early Access

We're building this
with schools, not for them

hushClass is in active development. We're looking for schools and teachers who want to co-build the future of classroom intelligence. Early partners will help shape the product and get priority access.

Building in Public

Questions We're Still Solving

Honest transparency about what we're figuring out. If you have insights on any of these, we'd love to hear from you.

How do we price this fairly?

Hardware + SaaS subscription? Per-classroom? Per-school? We need pilot data to determine real costs and value before setting prices.

What camera hardware is optimal?

Dedicated AI cameras vs existing CCTV? Single wide-angle vs multiple? Ceiling-mount vs wall? We're testing different setups in pilot classrooms.

How many languages can we support well?

Speech-to-text accuracy varies dramatically by language. We're starting with languages where STT models are strong and expanding as we validate quality.

Edge vs cloud processing?

Processing video on-device is better for privacy but limits AI capability. Cloud is more powerful but raises data concerns. We're building a hybrid approach.

Student privacy and consent at scale?

Face enrollment requires parental consent. Different jurisdictions have different rules. We're building consent management into the core platform.

How do we measure accuracy?

Is our engagement score actually accurate? How do we validate that ARIA's questions match real student confusion? We need classroom pilots to calibrate.