From Samarkand to Silicon Valley: how Sanjar Atamuradov is building the future of humanoid robots

Sanjar Atamuradov, a graduate of KAIST (South Korea) and a researcher at the Georgia Institute of Technology (USA), hails from Samarkand and is currently developing one of the most ambitious platforms in global robotics: Humanola, a startup based in Silicon Valley that redefines telepresence as the cornerstone for physical artificial intelligence. The company allows operators worldwide to control humanoid robots with minimal latency, transforming every interaction into training data for future physical AGI.

In an interview with IT Park Uzbekistan and The Tech, Sanjar shared his journey—from a young boy herding sheep under the vast night skies of Samarkand to the founder of a Silicon Valley startup that has secured $2.5 million to establish the groundwork for next-generation robotics.

Sanjar Atamuradov, Samarkand — founder of Humanola, LinkedIn

I was born in Samarkand, a city where childhood unfolds under a canopy of stars and where aspirations seem more attainable. From a young age, I was captivated by technology: I took apart anything I could find to understand how mechanisms functioned, eventually progressing to computers and robots. It was perhaps during this time that I first felt the ambition to one day create my own drones and robots.

After completing high school, I moved to South Korea to attend KAIST, one of the premier universities globally for engineering and robotics. There, I studied computer science with a specialization in robotics, focusing on the development of autonomous systems and drones. This experience marked the beginning of my journey: I recognized my desire to dedicate my life to humanoid robots and artificial intelligence.

While at KAIST, I aimed to immerse myself in AI and robotics as much as possible. Throughout the academic year, I dedicated most of my time to studying the theoretical dimensions of AI and robotics, and during breaks, I interned at various companies to apply my knowledge practically. In Korea, both summer and winter breaks span approximately 2.5 months, offering ample opportunity to engage in significant industry projects.

Initial steps in the field: computer vision and autonomous robots

Upon returning to South Korea, I interned at StoneLab Inc., where I played a role in developing a medical diagnostics application utilizing computer vision, which involved data annotation, neural network training, and model backend integration. Concurrently, I created an API for biometric authentication—my first system utilized by actual customers.

My next significant advancement was my work at Macroact Inc.

where I developed autonomous navigation for a quadruped robot — optimizing algorithms, fine-tuning behaviors, and configuring simulations in Gazebo (a software platform for creating virtual environments with robots and other entities, enabling developers to test robot behaviors and control algorithms without jeopardizing actual hardware). This served as a substantial foundation for progressing to more intricate robotic systems.

During my time at KAIST, I also engaged in research at AVE Lab and IRiS Lab. At AVE Lab, I created algorithms for self-driving vehicles. In IRiS Lab, my focus was on perception systems for autonomous vehicles — encompassing LiDAR processing (a digital representation of spatial data as a “point cloud” generated through laser scanning, known as Light Detection and Ranging) for lane detection. This phase provided me with a profound insight into how a robot perceives its environment.

Industry: from warehouse AMRs to quadruped robots

Following my research endeavors, I moved into the industry. At Digitrack Inc, I contributed to the development of autonomous mobile robots (AMRs) aimed at warehouse automation.

Subsequently, at Raion Robotics, I focused on locomotion for quadruped robots. This was a significant challenge — enabling a robot to walk steadily, adapt to various surfaces, and execute intricate maneuvers.

Studying and working in the United States: integrating engineering and science

My career path ultimately brought me to the United States, specifically Atlanta, where I immersed myself in an environment dedicated to advanced humanoid robotics research. I enrolled at the Georgia Institute of Technology (Georgia Tech) — recognized as one of the premier institutions for robotics and AI globally.

Studying in the United States marked a crucial phase in my development. I acquired extensive foundational knowledge in robot control, neural networks, perception systems, and autonomous behavior. I undertook courses in robotic manipulation, motion optimization, machine learning, and real-time algorithms — all of which are integral to contemporary robotics.

At the same time, I worked as a robotics engineer at Atlanta Ventures, where I developed mobile robots for security and inspection applications. This experience was where theory began to converge with practice — as classroom concepts seamlessly transformed into code, algorithms, and functional machines.

I subsequently continued my research at Georgia Tech, concentrating on scientific projects in the area of humanoid loco-manipulation — the capacity of humanoid robots to synchronize movements, walk while carrying objects, open doors, and execute complex sequential tasks.

During this time, I came to understand that large language models achieved significant advancements due to extensive internet data. In contrast, robots lack such an "internet" and are essentially deprived of data. I recognized that the only scalable method to produce high-quality training data is through teleoperation—robots managed by human operators. However, this approach had not been systematically addressed.

This realization led to the emergence of an idea that transformed my life: to create a platform that provides every roboticist with a telepresence tool and establishes infrastructure for physical AI data. Consequently, Humanola was founded in June 2025.

Currently, the enthusiasm surrounding robotics in the United States is remarkable—many view it as the next frontier of AI. Companies and research institutions are making substantial investments to tackle the universal robotics challenge, which is notably complex. However, recent advancements in AI have made the development of physical intelligence attainable. The field remains in its early stages: widespread implementation for physical labor automation has yet to occur. Humanoid robots operating in factories, warehouses, or homes in the U.S. are not yet a reality—but that moment is approaching rapidly.

The Problem Humanola Addresses

At present, robot training data is confined within individual companies and laboratories. Each entity gathers its own small datasets, with no sharing occurring, leading to stagnation in the sector. It resembles a scenario where every AI system in 2023 is trained on its own limited, localized micro-internet.

Prior to Humanola, each company was required to establish its own data collection infrastructure—an endeavor that was costly, slow, and inefficient. Many operated with restricted datasets, hindering progress. A cohesive ecosystem that integrates data and offers access simply did not exist.

Our team envisions a future where robots undertake hazardous, repetitive, or monotonous tasks, allowing humans to concentrate on creativity and strategy. The sole obstacle between the present and that future is the lack of data for physical AI. By tackling this issue, we can expedite the arrival of that future by several years.

I realized that the challenge lies not only in the fragmentation of individual solutions but also in the lack of an industry-wide platform that all market participants could engage with. The duplication of efforts and isolation of data hampers the entire sector. This insight became the driving force behind the creation of Humanola.

What Humanola Offers

Humanola is focused on developing a platform for remote robot control, along with a comprehensive infrastructure for the collection, processing, and analysis of physical data.

The platform is comprised of two essential components:

Consequently, businesses gain a robust tool that expedites robot development.

Humanola stands out as the sole independent platform that is not linked to specific hardware, providing a comprehensive solution — from robot control to complete data processing. Other alternatives necessitate that companies create their own infrastructure or lack such extensive integration.

Our clientele includes organizations utilizing robots in logistics, agriculture, and warehousing, along with developers and researchers who need swift access to high-quality data.

On challenges in the work

The most challenging aspect was achieving minimal latency over long distances while integrating various hardware. This demanded extensive engineering efforts at all levels — from the VR interface and network protocols to cloud infrastructure.

We accomplished this by creating our own real-time network infrastructure based on UDP protocols, incorporating forward error correction (FEC) and adaptive streaming, akin to leading VR games. For global control, we established regional edge servers (USA, Asia, Europe), enabling us to keep latency within 80–100 ms even across continents.

Hardware integration was achieved through a unified API for robots — abstracting joint control, sensor streams, and safety protocols, allowing any humanoid or manipulator to be connected with minimal configuration.

A significant advancement occurred when we merged private 5G networks in test environments with GPU-accelerated video encoding/decoding on both the operator and robot sides. This facilitated stable teleoperation with latency under 100 ms — for instance, from Tashkent to San Francisco. We demonstrated that global, scalable telepresence is indeed feasible.

The primary motivation to persevere stems from our belief in the mission and the realization that we are not merely creating a product, but laying the groundwork for the next generation of AI and robotics.

The technical breakthrough that enabled remote operation with sub-100-ms latency was pivotal — it marked the transition from concept to reality and allowed the product to enter the market.

What I take the most pride in:

We have successfully completed a seed round of $2.5 million from Link Ventures, Bad Ideas Fund, the Georgia Tech Fund, and Akmal Paiziev. Our initial clients have already started implementing the platform, and we are witnessing clear validation of both the significance of the problem and the value of our solution. Revenue metrics and user figures are not being disclosed due to the project's early stage.

We are backed by prominent venture funds, and our clients consist of companies that are making significant investments in robotics. The dynamics of adoption provide the most compelling evidence of effectiveness.

On Development in the USA, Korea, and Uzbekistan

The United States is at the forefront of developing the "brain" of robotics—AI and software systems that enable robots to think and act like humans. This is the key factor propelling the industry forward.

Korea boasts strong teams and consistently pushes technological limits. However, its scale is considerably smaller than that of the United States. Korean firms often concentrate on hardware and innovation but do not match the scale of the American market.

Uzbekistan is just beginning its journey in this field. Robotics in the country is still in its nascent stages of development.

Plans for the Near Future

Our goal is to integrate Humanola into all major humanoid robot platforms in the U.S. and worldwide, positioning ourselves as an infrastructure partner for industry leaders—Nvidia, Tesla, Google—in the development of physical AGI.

As part of our expansion strategy for the first year, we are focusing on large-scale deployments through robot manufacturers. We have already started hiring telepresence operators from Uzbekistan to manage robots deployed in the U.S. and other nations.

This model creates substantial economic opportunities—thousands of new jobs across Uzbekistan—while rapidly increasing the volume of training data collected. Operators can work from any location, robots can execute tasks anywhere, and data is returned to enhance models. This approach benefits the entire robotics sector, fosters job creation in emerging markets, and advances physical AI.

We aim to expedite the development of physical AGI—to transform robots into mass-market assistants and fundamentally change labor approaches. Our mission is to establish the infrastructure that democratizes the industry and makes intelligent robots accessible everywhere.

Advice for Entrepreneurs

Dream big and embrace risk. All significant accomplishments occur at the brink of failure—what truly matters is perseverance and consistency.