Building for India First
Every voice AI company starts with English. We didn't.
Why India
India has more phone calls per day than any other country. A huge portion of commerce — healthcare appointments, loan collections, e-commerce delivery, customer support — still happens on the phone. And a huge portion of those callers don't speak English.
If you want to build voice AI that actually reaches people, you have to work in Hindi, Tamil, Telugu, Kannada, Bengali, Marathi. The list goes on.
Most voice AI infrastructure treats Indian languages as an afterthought — plugging in a third-party ASR with poor accuracy and calling it "multilingual support." We think that's wrong.
What we built
We've invested specifically in:
- ASR models fine-tuned on Indian accents and code-switching (mixing Hindi and English mid-sentence, which is extremely common)
- TTS voices that sound natural — not robotic transliterations of English phonemes
- Latency optimized for Indian network conditions, including 4G connections with variable throughput
We currently support Hindi, Tamil, Telugu, Kannada, Bengali, and Marathi, with Gujarati and Malayalam coming next quarter.
The bigger picture
India is the hardest version of this problem. The accent diversity, the code-switching, the network conditions, the sheer volume of dialects — getting this right here means getting it right everywhere.
We're not "expanding to India." India is the product.
