While global machine-translation services have spent decades developing tools for English, Spanish, or Chinese, Turkic languages have remained almost entirely overlooked. Uzbek, Karakalpak, Kazakh, Turkish — rich, agglutinative languages — were long considered a “blind spot” in NLP.

 

It was precisely this niche that a young Uzbek engineer, Muhammadsaid Mamasaidov, entered in 2022 by launching a startup now widely recognized across the region — Tilmoch AI.

 

Tilmoch is an Uzbek technology company developing an artificial intelligence platform for translation and text correction in Uzbek, Karakalpak, and other Turkic languages. The startup grew out of the Tahrirchi project and is now considered one of the most promising AI services in the region. In 2025, Tilmoch raised $150,000 in investments from AloqaVentures, Yoshlar Ventures, and IT Park Ventures with a valuation of $1.5 million. The startup has also received several prestigious awards, including $50,000 from the mGov Award (2022) and $100,000 from the President Tech Award (2023). 

 

The platform has more than 130,000 users and over 50 corporate clients. Tilmoch supports a wide range of Turkic languages — from Uzbek and Kazakh to Turkish — reaching an audience of over 180 million people.

 

The startup successfully completed the AlchemistX Silicon Valley Residency program. During the Startup Showcase in the United States, Sergio Gor highlighted the startup’s significant potential and expressed confidence that such initiatives can strengthen international technological cooperation and unlock the talents of the region’s young generation. 

 

How it all began

 

The story of Tilmoch AI began long before the company itself — back in the founder’s university years. Muhammadsaid grew up surrounded by computers: his father was a programmer, and his interest in IT developed early. His first attempts included creating websites as a teenager. Later he considered pursuing medicine, but ultimately chose the technological path, graduating from an IT-specialized college and entering Inha University in Tashkent.

 

In his second year, students were asked to complete a self-selected project as part of the “Data Structures” course. Muhammadsaid chose to develop an Uzbek spell-checker — a task that proved far more difficult than expected. The first version was primitive, the instructor was dissatisfied, and the author himself admits he did not yet have a sufficient mastery of the language. However, this assignment became the turning point: the desire “to improve it” stayed with him and gradually evolved into research that uncovered important patterns of Uzbek morphology.

 

Today, more than 180 million people speak Turkic languages, yet major technology companies have historically overlooked them compared to the Romance-Germanic language group.

 

“Large services are not optimized for Turkic languages. They are trained on massive English corpora but are completely lost when encountering our morphology,” says Muhammadsaid.

 

Early challenges

 

Uzbek is an agglutinative language in which words can be very long and contain dozens of affixes. This linguistic characteristic makes Uzbek spell-checking a unique task: the system must account for approximately 200 affixes and their strict rules of combination. 

 

“At some point it became clear: the challenge was not the tools themselves but the absence of high-quality linguistic infrastructure for Turkic languages. We needed to build our own models, our own corpora, and our own approaches,” Muhammadsaid explains.

 

Despite the lack of electronic dictionaries, part-of-speech annotations, and insufficient data, Muhammadsaid continued collecting materials independently, studying morphology, and developing early algorithms. Only later, after the first success and funding, did a team of linguists join him, helping systematize affixation and create an almost complete dictionary of Uzbek word stems — a task that required three months. 

 

How Tilmoch works

 

It took five months to move from the idea to the first version of Tahrirchi. The startup trained — and continues to train — its models on regional data: literature, media content, local websites, parallel corpora (parallel corpora are collections of texts in which the original is aligned with translations in other languages), including texts dating back to the 1930s. The database now includes more than 40,000 books — an extensive resource for analyzing language across different eras. This enabled the development of a universal model capable not only of correcting errors but also analyzing style, tone, and historical linguistic variations.

 

The team views text generation and advanced translation models as future directions — not seeking to compete with global systems like ChatGPT, but instead creating specialized tools for Uzbek and other Turkic languages. Additional developments include systems for detecting toxic comments and various chatbot tools.

 

 

“We train models to understand language the way humans understand it. Without that, high-quality translation and proofreading are impossible — especially in Uzbek or Kazakh,” the engineering team notes.

 

For low-resource language pairs, the team uses hybrid approaches: fine-tuning models for Uzbek/Karakalpak/Kazakh, integrating spelling and grammar-check modules that consider local norms and spelling variations, and implementing post-editing. This enables Tilmoch to achieve higher quality for Turkic languages compared to mainstream global services.

 

Improvements are also supported by an internal evaluation system (automatic metrics), expert reviews and feedback from native speakers, A/B testing, and comparisons with major providers.

 

As a result, Tilmoch AI has evolved into a comprehensive ecosystem of language tools. Users can translate texts between Uzbek, Russian, English, Kazakh, Karakalpak, and Turkish. Beta versions of Chinese and Korean are already available.

 

The platform accounts for morphology and context — elements that global services often overlook. In addition to translation, Tilmoch offers advanced grammar and spell-checking — essentially an intelligent editor capable of analyzing sentence structure and correcting errors specific to local languages.

 

For businesses, the startup provides API integrations: companies can connect Tilmoch to their products, educational platforms, or internal localization systems. The objective is to ensure that language does not become a barrier in development, documentation, or customer communication.

 

The team emphasizes that this is only the beginning. Tilmoch is moving toward building a unified language engine capable of processing not only text but also voice, emotion, and speech styles. Plans include launching proprietary STT/TTS solutions, adding new Central Asian languages, and developing tools for corporate teams requiring large-scale content localization aligned with regional linguistic norms.

 

“We are not building a translator — we are building an ecosystem. Translation, editing, voice technologies, localization — one engine should handle all of it,” the team says.

 

More than 130,000 active users each month

 

Tilmoch AI currently has more than 130,000 monthly active users and 16,000 daily active users (DAU). Demand is especially high among civil servants, journalists, content teams, students, and professionals who require high-quality Uzbek or Kazakh text processing.

 

The corporate segment is expanding rapidly as well: more than 50 companies have already integrated Tilmoch into educational platforms, customer-service systems, localization applications, and internal administrative tools. The most active markets remain Uzbekistan and Kazakhstan, where the introduction of Kazakh-language support became a key driver of growth. The team is confident, however, that the potential extends further: Turkey, other Central Asian states, and the global Turkic-speaking diaspora.  

 

First investments and the creation of Tahrirchi

 

The first version of what eventually became Tahrirchi was built using personal funds. Attempts to secure government support were unsuccessful — yet a decisive moment came with the mGov Award 2022. The team developed a spell-checking Android keyboard weighing just 2 MB, which became one of the key factors in the project’s success. The project advanced through the selection stage, reached the finals, and received a $50,000 grant — the catalyst that enabled the launch of a full ecosystem.

 

 

After receiving this grant, the team expanded: linguists and annotators joined, helping build a comprehensive database of affixes and stems. T On this basis, the web editor was developed, followed by a Microsoft Word extension which — after lengthy and bureaucratically complex correspondence — Microsoft eventually published in its marketplace. Around 150 users actively use the extension today.

 

The startup also received $100,000 after winning the President Tech Award. Later, AloqaVentures, Yoshlar Ventures, and IT Park Ventures invested in the startup — the total amount raised at this stage reaching $150,000. Although the amounts may not compare with global venture rounds, for the Turkic NLP market they mean a great deal: they signal a growing confidence in a niche that had long been considered unpromising. 

 

The team notes that every award and each investment round represents not only funding but also validation that world-class technology can be built by addressing a local problem.

 

Tilmoch AI on the global stage: TechCrunch, AlchemistX, and C5+1

 

Tilmoch AI’s entry into the international arena marked a turning point in its development. The team first participated in the AlchemistX Silicon Valley Residency program, gaining access to mentorship, expert support, and a global network. This was followed by presentations and networking at TechCrunch — a platform where startups from around the world present their solutions to investors, corporations, and media.

 

 

A particularly significant milestone was the presentation of Tilmoch in Washington at the C5+1 conference — an event attended by representatives of the governments of the United States, Central Asian countries, and international institutions. Tilmoch was among three Uzbek startups invited to the Startup Showcase, which was attended by Sergio Gor, Special Representative of the President of the United States for South and Central Asia. 

 

Sergio Gor emphasized Tilmoch’s strong potential and expressed confidence that such initiatives could strengthen international technological cooperation and foster the talent of the region’s younger generation. 

 

 

“When you stand on stage in the United States and speak about the Uzbek language, people are surprised. But after the demonstration of quality, no questions remain,” the team recalls.

 

The team: people who love language and technology

 

Today, Tilmoch AI is a team of engineers, NLP researchers, developers, linguists, and product specialists united by a shared mission: to create the most advanced technological platform for Turkic languages. Within the company, technical expertise is valued alongside a profound respect for language. Many team members joined because they want Uzbek, Karakalpak, Kazakh, and other Turkic languages to be processed by AI with the same level of quality as English or Spanish.

 

The first core team was formed during the mGov Award stage — just three people: the founder, a data scientist, and a designer. This group built the earliest Tahrirchi model and Android keyboard from scratch: six months were dedicated to the model and three months on keyboard. After receiving the grant, the team expanded, and it became clear that the complexity of the product lay not only in linguistics but also in engineering — requiring a balance between linguists and technical expertise.

 

This led to the creation of two full-fledged teams: linguists and developers. Linguists were recruited through open calls; selected candidates completed test assignments and underwent training in the internal methodology. Their work proved fundamental. Annotators expanded the affix database, enabling the system to recognize nearly all words in the Uzbek language. Creating a complete dictionary of stems took around three months of meticulous work. 

 

The company conducts large-scale research activities: engineers experiment with model architectures, compile new corpora, and develop hybrid approaches for complex language pairs. A dedicated direction focuses on voice technologies: Tilmoch is actively seeking STT/TTS specialists, recognizing that the future of language platforms is multimodal. 

 

The key lesson: solve a real problem

 

The story of Tilmoch AI demonstrates that even a highly localized problem can lead to a large-scale regional product.

 

“If you see a pain point no one is solving — solve it. Do not wait for large companies to do it,” the founder says.

 

Tilmoch has become a clear example of how engineering persistence and an understanding of local markets can bring a project to the international arena.

 

Advice for founders in the region

 

The story of Tilmoch AI has become an important lesson for many young entrepreneurs in Uzbekistan and Central Asia. The team emphasizes that the journey begins not with an appealing idea but with a real problem felt by real users.

 

The founder advises not to hesitate in addressing local problems — they are often overlooked by global players, creating unique opportunities for startups. It is essential to test the product within the local community, engage with accelerators and grant programs, use the infrastructure of IT Park, and confidently enter international platforms.

 

The Tilmoch team is certain: if a product genuinely solves a problem and is built with high technical quality, it will find its users — regardless of how “small” the language or market may seem.

 

According to Tilmoch AI’s experience:

 

• focus on real pain points;

• test hypotheses;

• use accelerators and grants;

• do not hesitate to participate in TechCrunch, even if the product appears “too local”.

 

“There is room in the global market for those who build products for real people, not only for speakers of major languages,” the team believes. 

 

Tilmoch — a member of IT Park Uzbekistan

 

Tilmoch AI is more than a translation tool. It represents an effort to build technological infrastructure for tens of millions of people whose languages have long remained outside the focus of global AI. Judging by the team’s progress, this mission is already becoming a reality. 

 

Today Tilmoch AI is one of the clearest examples of how a local technological initiative can grow into a regional and international project while maintaining a deep connection to culture and linguistic identity.

 

The startup has demonstrated that innovation can emerge not only in major language ecosystems but also in areas previously considered “invisible” to global AI. 

 

As a member of IT Park Uzbekistan, the startup benefits from a number of tax incentives that are critical to the financial sustainability of a young company. Exemption from personal income tax and reduced rates for certain categories of payments significantly reduce operational costs. This enables the team to allocate most of its budget not to administrative overhead but directly to product development: expanding the team of engineers and linguists, experimenting with models, and creating new functions and integrations for businesses. 

 

Website:

https://tilmoch.ai

Telegram:

https://t.me/tilmochai

Telegram bot:

https://t.me/tahrirchi_uzbot

Instagram:

https://www.instagram.com/tilmoch_ai