
Enterprise Workforce Management | Real-Time Tracking & Automation
A centralized workforce management platform that streamlines onboarding, approvals, resource allocation..

An evidence-based engine that compares classroom transcripts against the official syllabus and reports, topic by topic, how thoroughly each subject was actually taught.
Our client is an education technology company serving multilingual classrooms across India. Their platform depends on accurate lesson records, but classrooms speak in a mix of English and regional languages, often switching between them within a single sentence. Any transcription tool had to work across that mix, run on modest hardware already installed in schools, and never send student audio off site.









Syllabi are often written in English while lessons are taught in a regional language
School administrators needed proof, not just a score, for every coverage decision
The system had to be benchmarked and proven reliable before being trusted for reporting
Checking syllabus coverage by hand across many classrooms and subjects was slow and inconsistent
Source syllabus documents arrived in several different structures and formats
We built a pipeline that transcribes audio in short streaming chunks while the class is still in progress, using voice-activity detection so processing time is spent only on actual speech. Once the class ends, a final reconciliation pass merges the streamed chunks into a single clean, time-stamped transcript.
Development happens on standard GPUs, but deployment happens on low-power edge devices in the field. We built the system so the exact same codebase runs on both without modification, removing an entire class of deployment bugs.
Each spoken segment is automatically labelled with its language and transcribed in its original script. Nothing is translated, so the transcript reads exactly as the class was taught, including natural code-switching between English and a regional language.
Every stage, from audio capture to final transcript, runs on the client's own hardware. No student data leaves the premises, and the system keeps working even when the internet does not.
Every coverage score comes with a time-stamped trail showing exactly where and how a topic was taught.
Schools can compare an English syllabus against lessons taught in a regional language with no manual translation step.
Administrators get a simple Red, Amber, or Green view per topic, with the ability to drill into the underlying evidence.
The system was benchmarked against labelled test cases and met an accuracy target of eighty-five percent or higher.

By pairing structured syllabus parsing with hybrid search and an auditable evidence ledger, this system turns curriculum coverage from a manual guess into a defensible, evidence-backed measurement, one that works across languages and holds up to scrutiny.
Explore other projects we've delivered

A centralized workforce management platform that streamlines onboarding, approvals, resource allocation..

An AI-powered phone ordering system that takes orders by voice across multiple restaurant locations an..

AI-powered aggregation and verification system that continuously collects global sanctions databases and..