In Conversation with Sardor Davlatov of NavAI

How did you end up in voice?

A few years back, Sardor Davlatov called the DMV to book an appointment for a friend. The agent on the line sounded completely human, smooth, and natural, with no hesitation. At the end of the call, it told him he'd been talking to an AI the whole time.

That moment changed the direction of his career. If voice AI could pull this off in English, why couldn't it do the same in Uzbek? There was no dataset to fine-tune, no existing model to lean on. Just a low-resource language and a region that text-first tech had quietly skipped over.

That question became NavAI.

We sat down with Sardor to understand the inner workings of NavAI and how he plans to navigate the voice AI landscape.

What's your struggle or moment of joy with voice?

The hardest part has been the language itself. Uzbek is a low-resource language: no clean datasets, no off-the-shelf models, nothing to build on. We had to train on thousands of hours of real speech from scratch. But that's also where the joy is. The first time our AI handled a real customer call end-to-end in Uzbek, and the caller spoke to it as naturally as they would to a person, was unforgettable. Seeing people in our region finally served by technology in their own language is what keeps me going.

Where do you think the voice is going?

Everywhere. I think voice becomes the default way we interact with technology. Just as we moved from typing commands to tapping screens, we'll gradually shift from keyboards to simply speaking. Apps will let you navigate and take actions by voice; calling a business will mean talking to an AI that genuinely understands you, in your own language, 24/7. Typing won't disappear, but voice will become the most natural interface. And the regions and languages that text-first technology left behind will finally be included.