Case study

Sarthi: live conversations turned into operational intelligence

A web and desktop platform connecting live call audio, transcription and AI analysis, so teams can work with summaries, detected conditions and flags in one operational interface.

Industry: Conversation operations
Services involved: AI development Software development Integration & automation

The problem

Phone conversations contain valuable information, but reviewing them individually does not create a systematic view. Sarthi needed the complete path from live audio to structured intelligence and human action.

The shape of the result

Audio → actionable insight

Inside the product

What we built

Live audio pipeline

Call audio enters the backend and moves through transcription into structured analysis.

Intelligence beyond text

Summarisation, detection and flagging transform transcripts into reviewable information.

People in control

Web and desktop interfaces bring the intelligence into a human review and action loop.

1Live call audio
2Transcription
3AI analysis
4Human review

The decisions behind the build

01
Use the backend as the orchestration layer between live call audio and the intelligence pipeline.
02
Separate speech-to-text from reasoning: Groq Whisper Large V3 Turbo handles transcription; Claude Haiku 4.5 analyses the text.
03
Turn transcripts into summaries, detected conditions and flags rather than stopping at raw text.
04
Bring the outputs into web and desktop interfaces where people can review and act.
05
Test the full pipeline, observe actual behaviour and revise individual components without rebuilding the system.

What we delivered

A connected conversation-intelligence platform spanning call processing, transcription and AI analysis.
Summaries, detection and flags surfaced for operational review.
Web and desktop experiences around the same underlying intelligence pipeline.
Speech recognition and reasoning kept independent so components can evolve separately.