AI Companies in Uzbekistan Hiring in 2026: Who's Building, Who's Paying, and How to Get In

“There is no AI work in Uzbekistan” was a defensible claim in 2022. In 2026 it is simply wrong. The state has an AI strategy running to 2030 with a projected ten billion dollars of GDP impact and over 860 million euros of targeted foreign investment. DataVolt has broken ground on a green data center inside Tashkent’s IT Park, with memoranda pointing at campuses beyond 250 megawatts. A tax-free regime for large AI and data-center projects has been carved out to pull in more. And beneath the government announcements, an actual job market has formed. Here is who is in it.

The speech and language startups: Uzbek as a moat

The most distinctive AI work in the country is happening in voice. Aisha AIbuilds Uzbek-first voice agents, TTS and STT models, and call analytics across Uzbek, Russian, and English — production systems with over a million voice interactions behind them. Muxlisa AI ships speech-to-text and text-to-speech tuned for Uzbek and its dialects. The UzbekVoice / mohir.aiecosystem grew from a crowdsourced corpus project into working speech APIs. For an ML engineer, these teams offer something Silicon Valley cannot: low-resource language problems where your model is the state of the art for thirty million speakers, not a fine-tune of someone else’s leaderboard.

The fintechs: the biggest ML teams nobody calls AI companies

Uzum— the country’s first unicorn — runs recommendation, search, credit scoring, and fraud models across an e-commerce and fintech super-app. TBC Uzbekistan, with Payme in its group, has moved from pilots to production AI agents handling payment-reminder calls. Clickand the banks are on the same road. These companies rarely brand themselves as AI employers, yet they hire more data scientists and ML engineers than anyone else on the local market — and they interview like fintechs: expect scoring metrics, class imbalance, and “what happens when your model blocks a legitimate payment” rather than transformer trivia.

The global firms: EPAM, Exadel, and the delivery route

EPAM hires ML and AI engineering roles up to team-lead level, remote from anywhere in Uzbekistan, plugged into international client projects. Exadelrecruits AI-focused data engineers across its locations including Uzbekistan. The work is client-driven rather than research, but it comes with two things local startups struggle to offer: international-scale production systems and English-heavy teams — the exact experience that later unlocks EUR-denominated remote roles. Their interviews follow the global template: coding rounds, ML system design, and behavioral rounds in English.

The state layer: strategy money becomes jobs

The 2030 strategy is not just a decree. It funds AI laboratories with universities, a dedicated Center for the Development of AI and the Digital Economy, and infrastructure spending that has already put tens of millions of dollars into compute. Data centers being built by DataVolt and the Karakalpakstan tax-free zone are bets that training and inference capacity will live in-country. State-adjacent AI work — national language models, government service automation, health and education pilots — is growing into a genuine employer for engineers who want impact measured in millions of users.

What the market pays and how to get in

ML roles top our salary table: roughly 12–18 million UZS for juniors, 25–38 for middles, 45-plus for seniors, with public estimates putting the Tashkent average around 21 million — the highest-paid engineering discipline in the country. Getting in is less mysterious than it looks. The voice startups want to see one real model you trained and can defend. The fintechs want SQL, metrics thinking, and business sense. The global firms want clean coding rounds in English. All three formats are rehearsable: run mock interviews for the ML or data role you are targeting, in the language the real loop will use, until the questions feel familiar. The market is small enough that a strong interview gets remembered — and talked about.