Indian Startup Builds Voice AI for 220 Languages

Most voice AI systems fail the moment someone switches languages mid-sentence or speaks with a regional accent. For most of Asia, that makes the technology nearly useless.

Ritu Mehrotra and Sourav Bandyopadhyay saw this firsthand while building a mental health platform. Their users mixed Hindi and English freely, spoke dozens of dialects, and needed their sensitive health data to stay private. The voice technology available couldn't handle any of it.

So they built their own. Shunya Labs now creates voice AI infrastructure designed specifically for how people actually talk in multilingual countries.

The difference shows up immediately in real conversations. When a customer calls their bank and switches from English to Tamil mid-sentence, Shunya's system keeps up. When a doctor dictates notes using technical terms over background noise, the transcription stays accurate.

The company built everything from scratch: speech recognition, processing systems, and voice generation. The entire platform runs on regular computer processors instead of expensive specialized chips, cutting deployment costs to about one-twentieth of competitors.

Shunya covers more than 220 languages, including 55 from India. Adding a new language takes around 100 hours of training instead of the industry standard of 10,000 hours. Each business customer gets models trained on their specific vocabulary, whether that's pharmaceutical terms, banking products, or telecom plan names.

The technology works offline too. It runs on phones, in cars, or on a company's own servers without needing internet connectivity.

Major companies across banking, healthcare, telecom, and automotive industries now use the platform. Shunya recently released an open-source translation system called Vāķ that handles real-time translation across 55 Indian languages with less than 1.5 seconds of delay.

The Ripple Effect

The impact reaches far beyond smoother customer service calls. Doctors can now dictate patient notes in their preferred language while the system automatically structures information for medical records. Rural patients can access healthcare services in their native dialect. Bank customers can verify their identity or check balances without struggling through English-only menus.

For the millions of Indians who naturally blend languages in conversation, technology is finally catching up to how they actually communicate. Revenue has doubled month over month for the past three months as more companies realize their customers need this flexibility.

Every conversation in a person's own language becomes a small moment of dignity and inclusion.