Unitlab: How a team from Tashkent is building an AI business for the U.S. market

Unitlab founder Shohruh Bekmirzaev shared with Spot the reasons behind registering the company in the United States, how former colleagues contributed to product enhancements, his approach to hiring global specialists, and the amount of funding he has secured from venture capital firms.

Shohruh Bekmirzaev is a former artificial intelligence engineer who spent several years in technology firms in South Korea before leaving a secure position to develop his own product in Uzbekistan. This endeavor resulted in the establishment of Unitlab AI, a platform focused on data annotation.

The concept for the startup emerged from personal experiences. While engaged in AI work, Shohruh faced a challenge common to many AI teams: the data annotation process was slow, costly, and heavily reliant on manual labor.

The annotation process is crucial for training artificial intelligence. Without it, AI lacks comprehension of its inputs: an image is merely a collection of pixels, and text is simply a string of characters. To train a model, humans must manually clarify the content of the data—such as delineating an object within an image and assigning a label. This method demands considerable time and resources.

Consequently, an increasing number of companies are shifting towards automating data annotation through AI. Unitlab is one such solution. In a brief timeframe, the startup has penetrated the international market and garnered interest from major corporations, including Samsung. Currently, over 2,000 teams from 40 countries utilize the platform, including teams from the U.S., South Korea, Japan, and India.

In an interview with Spot, Shohruh Bekmirzaev elaborated on his decision to register the company in the U.S., the role of former colleagues in refining the product, his global hiring strategy, and the total investment the company has attracted from venture funds.

How It All Started

I completed my bachelor’s degree at Tashkent University of Information Technologies, and in 2018, I obtained a master’s degree in computer science from Kumoh National Institute of Technology in South Korea, specializing in artificial intelligence and deep learning.

Upon graduation, I began my career as an AI engineer at the South Korean firm Lululab, where I filed five patents in the field of artificial intelligence, two of which were registered in the United States. Subsequently, I joined Mathpresso, where I led the development of the core AI engine for over three years.

Although I had a stable job, I always aspired to create my own product.

In 2022, I made the decision to resign and return to my home country, fully committing to the establishment of a company aimed at addressing a recurring issue I faced: data annotation for AI. At that time, this process was characterized by its slowness, high costs, and inefficiency.

During my tenure at various tech companies, I recognized that the primary obstacle in AI development was not the training of models, but rather the substantial time and resource expenditures linked to data annotation. We resolved to transform this frustration into a source of innovation by automating the annotation process and removing unnecessary manual tasks.

Thus, Unitlab AI was founded.

Initially, I recruited two engineers from my professional network—Ahror Baratov and Shahzod Uralov. As a small team, we dedicated ourselves to developing the first version of the platform. After they showcased a remarkable level of expertise and commitment, they became co-founders of the company. Today, the three of us are collaboratively advancing Unitlab AI.

For over three years, until we secured our first external investments, I completely self-funded the company: I invested more than $170,000, refrained from taking a salary, and concentrated solely on the product.

The initial version of Unitlab was quite basic—a fundamental image annotation tool. Our objective at that stage was straightforward: to assist AI teams in preparing training data more quickly and cost-effectively.

Hypothesis Testing and First Customers

To ensure we were addressing a genuine problem, I contacted former colleagues in South Korea and invited them to test our platform. We received comprehensive feedback from them, which led us to invest further in the product: we brought on two additional engineers and a designer to enhance functionality and UI/UX significantly.

Subsequently, I began reaching out directly to our target audience via LinkedIn, inviting them to test the product. Some of these individuals became our initial paying customers.

We also prioritized technical content: I enlisted a technical blogger to oversee our corporate blog and later expanded the author team to four, including specialists from overseas. We shared these articles within targeted AI/ML communities on LinkedIn and Facebook—this became our first reliable user acquisition channel.

As time progressed, our technical articles began to rank on Google, which greatly improved our SEO. An increasing number of teams started discovering Unitlab AI through search queries relevant to their work tasks.

We also incorporated an automated feedback collection system into the product, allowing us to enhance the platform continuously based on genuine user input.

Our initial active users comprised AI startups and university labs. In the early stages, we encountered challenges—we misjudged the intricacies of onboarding and created an overly complicated interface. Nevertheless, each error provided us with valuable insights. We streamlined the interface, bolstered automation, and enhanced the overall user experience.

When we observed that over 1,500 teams had joined the platform organically and companies like Samsung began requesting tailored demos, it became evident that the issue we were addressing was widespread—and it was time to expand.

Registration in the United States

Although the Unitlab team is based in Tashkent, the company is officially registered in the United States. We made this choice primarily to access the global payment infrastructure, particularly Stripe, which is crucial for engaging clients in the U.S. and Europe.

Today, anyone can establish a company online in nearly any country, gain access to local banking systems, and attract customers globally through effective marketing—without needing to be physically present.

The entire registration process was entirely legal. I submitted an online application via the official Delaware state portal, completed identity verification, and received approval. Subsequently, we opened a U.S. bank account, integrated Stripe, and began accepting payments from international clients.

Currently, more than 2,000 teams from 40 countries utilize the product. These primarily include companies from fintech, retail, e-commerce, EdTech, and HealthTech sectors. The platform is also trusted by over 39 universities, including Carnegie Mellon University and the University of Mississippi.

Our strongest presence is in the U.S., Europe, and Asia (Japan, Korea, India, China). Each market has distinct requirements: in the U.S., scalability and integrations are vital, while in Asia, stringent compliance and on-premise solutions take precedence.

The U.S. market is highly competitive, featuring players such as Scale AI, Labelbox, and Roboflow. However, unlike most of our competitors, we emphasize automation over crowdsourcing. This strategy enables us to accelerate annotation by 15 times and cut costs by five times.

We address a critical challenge in AI projects—the reliance on manual annotation, which hampers development and inflates budgets.

Users have the ability to upload data in various formats—images, text, or audio—utilize auto-labeling, set up review pipelines, and incorporate their own models (BYOM). Annotation management, quality assurance, and automation are all conducted within a unified interface.

The foundation of our product lies in our auto-annotation AI engine and adaptable workflows. Clients can initiate large-scale projects in mere seconds, automate validation processes, and integrate their own models. For instance, a hospital might incorporate its medical model, or a bank could integrate an anti-fraud model into our pipeline—an aspect I take particular pride in.

We also provide enterprise-level features: on-premise deployment, BYOM, and support for medical imaging (DICOM). This positions us as especially appealing to enterprises with stringent compliance needs and specialized applications.

How We Secured Investment

Unitlab AI was initially built using a bootstrapping approach. After approximately three years of achieving sustainable traction, we began reaching out to investors on our own—sending emails, arranging meetings, and consistently pitching our concept. This endeavor spanned several months. Following each call and demonstration, we provided updates, which helped to build trust.

Concurrently, we applied to the competitive accelerators 500 Global and Startup Wise Guys. We participated in several meetings, each lasting over an hour, where we thoroughly discussed the product, market, traction, and team. After the final meeting, I received an invitation to join the cohort with investment.

Investors were persuaded not only by the size of the data annotation market (approximately $12 billion) but also by our rapid execution, genuine customer demand, and the team's unwavering dedication.

In total, we secured $610,000 from four venture funds:

These funds enable us to enhance sales, recruit essential personnel, and develop our infrastructure. Over the next year, we aim to increase ARR by 5–10x and penetrate new markets.

Our current objectives include achieving key metrics, boosting revenue, expanding enterprise partnerships, and forming strategic alliances. Following these goals, we plan to initiate a new investment round.

We operate on a freemium SaaS model, which means the platform is accessible to everyone and offers a free basic plan. Advanced features can be accessed through paid plans ranging from $99 to $195 per month per team. Custom pricing, on-premise installations, and extended support are available for enterprise clients.

Our objective is to achieve $1 million in ARR by the conclusion of 2026.

Growth and Challenges

Initially, we marketed our product to small teams and individual users, gradually transitioning to partnerships with companies. Currently, our main emphasis is on enterprise clients.

The enterprise sector presents the greatest challenges. Selling to large corporations differs significantly from onboarding standard SaaS users. The decision-making process is prolonged, trust needs to be established, and there are multiple levels of approval and compliance checks involved. For a startup based in Uzbekistan without an extensive sales team in the U.S. or Europe, this poses a considerable challenge.

One of the primary obstacles has been the labor market. While Uzbekistan has a wealth of skilled engineers, there are very few professionals with experience in B2B sales and launching SaaS products on an international scale. Consequently, we had to seek such talent abroad and recruit sales and marketing managers with global experience.

Initially, we conducted our own search by posting job openings and interviewing candidates. Simultaneously, we collaborated with recruitment agencies that assisted with preliminary screenings. Our approach was straightforward: we fully vetted candidates before paying the agency’s fee.

This process enabled us to hire specialists from other countries—individuals who had experience with global SaaS products, understood how to construct sales funnels, engage with enterprise clients, and manage extended sales cycles.

As a result, our team became hybrid: a robust engineering foundation in Uzbekistan complemented by commercial expertise from abroad. This synergy allowed us to simultaneously develop the product and establish a sustainable global go-to-market strategy.

Currently, our sales department consists of five individuals from Europe (Serbia, Kosovo, Spain) and the Philippines, all of whom work remotely on a full-time basis. Base salaries range from $2,500 to $3,500, in addition to monthly commissions of $1,000–1,500.

We formalize contracts with each employee, and salaries are disbursed through the company’s U.S. bank account, which is the most convenient option for our international team.

Our work processes have been established gradually. Engineers operate from the office, while remote employees connect through regular calls, task trackers, reports, and defined KPIs. This structure has allowed us to maintain speed, transparency, and control, even with a distributed team. In total, the company now employs 14 individuals: engineers, AI engineers, and sales and marketing specialists.

Our target markets are the U.S., Western Europe, and China.

These areas are where AI is advancing most rapidly, and where we observe the highest potential for revenue growth.

We are already established in the U.S. market, but we are not focusing on the mass market. Our emphasis is on enterprise clients and large organizations, including those in the aviation and defense industries.

The primary challenge lies in trust and reputation. Competing with American firms while operating outside of Silicon Valley necessitates robust case studies, references, and demonstrated product reliability. Additionally, compliance requirements add complexity: larger clients are increasingly demanding on-premise or hybrid deployments.

To engage with such clients, it is essential to adhere to international security standards, such as ISO/IEC 27001, which encompasses risk management, access control, data protection, and business continuity.

Our long-term objective is to evolve into a global data automation platform with regional offices or partners in every significant AI hub.

In the future, we intend to advance in three areas: geographic expansion, the introduction of new modules (medical imaging, video annotation, agent-based AI workflows), and the development of a BYOD-GPU computing platform that will enable clients to connect their own GPU servers for AI model training.