Developer Builds Multilingual AI Voice Agent for India in 10-Day Challenge
A developer participated in the '10 Days of AI Voice Agents — VoiceForBharat Edition' challenge, building an AI voice agent named Sadie designed to support natural spoken interactions in India. Sadie was built using Deepgram for speech-to-text, Google Gemini as its reasoning core, Murf Falcon for text-to-speech, and LiveKit for real-time audio, with added support for Hindi-English code mixing. The agent was equipped with persistent memory via a SQLite database, external dictionary tool integration, outbound phone call capability, and a human escalation system using Discord webhooks. Specialist agents for tasks like grammar assistance were also incorporated, alongside a real-time analytics dashboard built with Next.js. The project highlighted that building a capable voice agent requires far more than combining a language model with a speech API, involving telephony, memory, tooling, and human handoff systems.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in