Case Study · INFOTAINMENT · UK/EU

Kaft: Multi-agent LLM news podcast

Production AI audio with dual hosts and multi-language.

The story

What Kaft needed, and what we built.

Founders Tristan, Otis, Victor (Columbia, LSE) wanted to build an AI-powered news podcast app that customises content to user preferences, but production-grade, not a demo. KUMO built the full app, multi-agent LLM architecture, dual-host TTS, multi-language support, dynamic content generation.

What we delivered

Six areas of production work.

Multi-agent LLM orchestration

Specialist agents for research, synthesis, scripting, and audio assembly working through structured handoffs.

Dual-host TTS

Two natural conversational voices with prosody and turn-taking that sound like real podcast hosts.

Multi-language support

European, Asian, and American languages handled across research and audio generation.

Dynamic content generation

Personalised news content adapted to listener preferences, topics, and depth signals.

App Store deployment

Live mobile app with a production audio pipeline and offline-ready playback.

Editorial + AI workflow

Designed so the AI handles synthesis while editorial guidance keeps content shaped to brand and accuracy.

Highlights

Project highlights.

App Store Live
Multi-agent Production architecture
Multi-language Supported
Dual-host TTS Natural audio

Tell us what you're solving for.

We'll listen first, ask the right questions, and follow up with a clear proposal.