QazCode, one of Central Asia’s fastest-growing technology companies, is accelerating its global ambitions as it expands beyond Kazakhstan and deepens its footprint in Qatar and the broader Middle East. In an interview with The Peninsula, CEO Oleksii Sharavar outlined how a former telecom IT division evolved into a regional AI powerhouse now driving innovation across multiple industries.
Sharavar explained that QazCode was born from a strategic shift within a major Kazakh telecom operator: “Simply being a telco is not enough. You need to offer better products and improve the customer experience. That’s why we need to become an IT company.” Today, QazCode employs 750 people—half of them software developers—and operates across all VEON markets in Central and Southeast Asia as well as the UAE.
Building AI for Languages Big Tech Ignores
The company’s pivot to AI began when customer interactions surged beyond one million per month, exposing the limitations of global language models in understanding Kazakhstan’s multilingual population. Clients frequently switch between Kazakh, Russian, and Uzbek mid-sentence—patterns no major AI company had prioritized.
Faced with this gap, QazCode developed KazLLM, a locally trained multilingual model built on 150 billion tokens. The dataset—assembled from books, research, and academic archives with support from the Ministry of AI and leading universities—has no equivalent online. The model, published on Hugging Face for non-commercial use, understands local dialects, cultural context, and mixed-language speech rare in global systems. It was trained entirely on local infrastructure, ensuring data sovereignty—an increasingly critical requirement for governments and enterprises.
KazLLM is now being deployed across education, healthcare, enterprise support, and lifestyle applications, demonstrating the scalability of locally built AI.
Agentic AI is the Next Frontier
QazCode’s next major innovation is Aventa, an agentic AI platform designed for both enterprises and consumers. The company is developing industry-specific agents, including:
- an AI doctor assistant for improved diagnostics
- agriculture agents for farmers in Pakistan and Bangladesh
- corporate agents handling HR, compliance, legal tasks, financial analysis, and software development
Sharavar’s vision is to create AI tools that feel less like apps and more like “digital friends” that understand user histories without overwhelming them.
Middle East: A Strategic Growth Hub
The company is actively expanding into Qatar and the wider MENA region, which Sharavar credits for its pro-innovation environment, strong infrastructure, and investment appetite. Unlike Europe, he noted, the Middle East encourages startups to build first, regulate later—a “risk-accepting approach” he believes fuels faster innovation.
At MWC Doha 2025, QazCode signed a major partnership with a US sovereign AI firm to build low-cost, locally deployed LLMs for Bangla, Urdu, Ukrainian, Kazakh, and Uzbek—expected to reach more than 70 countries. Another agreement with JazzCash aims to develop a specialized AI customer service agent for Pakistan’s leading financial app.
Local Infrastructure, Local Data, Local Control
Sharavar emphasizes three differentiators that give QazCode an edge over global competitors:
- Local infrastructure enabling fast, low-cost AI inference
- Unique, curated datasets unavailable online, gathered through academic and government cooperation
- LLM-agnostic flexibility, allowing Aventa to run on Google, GPT-based, DeepSeek, or KazLLM models
The company also stresses ethics and safety, avoiding biases baked into internet-scraped datasets, and ensuring all sensitive data stays within national borders.
From Kazakhstan to the World
As interest grows from governments and enterprises across continents, Sharavar sees QazCode as proof that AI innovation is no longer the exclusive domain of global tech giants. Countries often overlooked by the global tech ecosystem can now build their own sovereign AI models tailored to local languages and cultures.
QazCode already has contracts in 15 countries and aims to expand to more than 70. Sharavar expects the firm to become an “AI-native operator” across its markets within three years, delivering digital services powered entirely by agentic AI and localized LLMs.
For Kazakhstan, QazCode’s rise marks the beginning of a new era—one where regional languages, cultural context, and national datasets shape the next generation of AI innovation.
