GenGenAI: "Enhancing AI Efficiency and Accuracy with Realistic Synthetic Data"

GenGenAI provides synthetic data generation solutions for AI model training. Unlike traditional generative AI services, their solutions generate synthetic data that reflects the characteristics of real camera sensors, making it similar to actual data. This technology allows AI developers to acquire the necessary data for the development of models quickly and efficiently, thereby saving on time and costs while improving the effectiveness and accuracy of AI training.

GenGenAI currently offers two solutions: "GenGenData," which provides synthetic data; and "GenGenVision," which develops and delivers AI models trained with GenGenData. These solutions are already being used across various industries, including the automotive and defense sectors. By making it easier to acquire data for different environments and special scenarios, the company has received positive feedback. GenGenAI is planning to officially enter the global market starting with CES 2025 in Las Vegas, solidifying its position in the field of synthetic data technology.

We spoke with Ho-jin Cho, CEO of GenGenAI, to learn more about the company and its groundbreaking synthetic data solutions.

Ho-jin Cho, CEO of GenGenAI / Source: iT DongA


Solving AI Training Data Challenges with Synthetic Data
iT DongA: Hello, Mr. Cho. Could you please introduce yourself?
Ho-jin Cho: Hello, I’m Ho-jin Cho, CEO of GenGenAI. I earned my Bachelor's and Doctoral degrees in computer science from POSTECH (Pohang University of Science and Technology). During my graduate studies, I focused on video quality improvements and restorations. I also served as a doctoral intern with the Adobe Photoshop development team in the United States. Afterward, I joined a domestic tech startup as an early member, where I spent 7 years and 4 months researching and developing vision AI technology. During that time, I developed technologies designed for real-time text detection and recognition for use in wearable devices, and real-time object detection and recognition for automobiles. In January 2022, I founded GenGenAI.
iT DongA: What motivated you to establish GenGenAI?

Ho-jin Cho: Developing AI models for autonomous driving requires millions of images or videos to be captured under diverse weather conditions, seasons, times, and locations. Certain types of data, such as wildlife appearing on roads or global road signs, are essential yet also difficult to obtain. I realized that the synthetic data generated through generative AI could address this problem. Observing how AI developers invest a significant amount of time, costs, and manpower in data collection and labeling further solidified my decision to start the company.

iT DongA: Can you tell us more about GenGenAI?
Ho-jin Cho: GenGenAI creates synthetic data that mimics the images captured by actual cameras, considering their sensor characteristics. This enables AI models to be trained more effectively and accurately, while also reducing the time and costs required for data collection and improving productivity.

We currently provide data for the development of autonomous driving and advanced driver assistance systems (ADAS), as well as driver monitoring data, target recognition and friend-or-foe identification data for the defense industry, and data for unexpected and exceptional scenarios involving CCTV systems. Our company name, "GenGenAI," which stands for "Genuine Generative AI," signifies our goal to create true generative AI.

Synthetic data for roads, drivers, road signs, and defense industry applications, generated by GenGenData / Source: GenGenAI


Creating Synthetic Data Similar to Real Data
iT DongA: Can you elaborate on GenGenAI’s technology and solutions?
Ho-jin Cho: We currently offer two solutions: "GenGenData," which provides synthetic data for AI training; and "GenGenVision," which delivers AI models trained with the synthetic data.

GenGenData is a solution that quickly generates the synthetic data that customers need. Our core technologies include ▲Generation: Creating specific scenarios based on existing images ▲Transformation: Altering the weather or seasons to generate new data ▲Composition: Adding separate images to create composite data.

For instance, it’s challenging to collect data on the wildlife that can appear on roads, let alone under various times, seasons, or weather conditions. With GenGenData, we can easily create such scenarios, adjusting the positions, sizes, and environments. We can also generate data for natural disasters like floods or explosions.

The realism of our synthetic data is another key advantage for us. Most of the existing solutions provide rendered graphic data from digital twin environments, which differ significantly from the actual camera-captured images. AI trained with such synthetic data tends to perform 20–30% worse in real-world scenarios compared to when it is trained with real data.

In contrast, our technology incorporates camera sensor characteristics, such as the color, noise, field of view, and lens refraction, to generate data that is almost identical to real camera footage. The AI models trained with our synthetic data achieve performance levels that are comparable to training with real data. In addition, increasing the data volume can further enhance the performance, as we can generate unlimited data.

GenGenVision builds upon GenGenData by training AI vision models with the generated synthetic data. Customers can quickly adopt these validated models and focus on their core areas of business.

Explaining synthetic data solutions / Source: iT DongA


Expanding Across Various Fields and Global Markets
iT DongA: How has your business been progressing so far?
Ho-jin Cho: We are currently receiving requests for synthetic data from the automotive, defense, and security sectors. In the automotive field, we mainly provide data for training AI models designed for autonomous driving. In the defense industry, we are collaborating with military organizations, research institutes, and companies. For example, we’ve conducted PoC projects to create AI training images based on satellite, aircraft, and drone footage. In the security sector, we supply synthetic data for various unexpected scenarios and emergencies in this field.

We are also generating revenue in the global market. Although we are not actively engaging in marketing or sales, we have received collaboration requests, some of which have led to PoC agreements. After these projects, we are planning to formalize some partnerships next year.

Recently, we received the CES 2025 Innovation Award in the AI category, which demonstrates the global recognition of our technology. We were also selected as one of the "100 Innovative Defense Companies" by the Defense Acquisition Program Administration, through which we will receive up to 5 billion KRW over three years in order to develop core AI and data technologies for the defense sector.

As for funding, we secured a seed investment from Naver D2SF and DSC Investment in April 2022. In September 2023, we attracted pre-Series A funding from Smilegate, DSC Investment, Korea Development Bank, Stonebridge Ventures, HL Mando, and Innowireless, leading to a total accumulated investment of 5.3 billion KRW.

iT DongA: How has SBA supported GenGenAI?
Ho-jin Cho: SBA supported us through an open innovation program with Hanwha Systems, where we validated the applicability of our synthetic data for the defense sector. This program allowed us to refine our solution based on expert feedback and laid the foundation for our expansion into the defense market. SBA also supported our participation at CES 2025 by providing us with a booth.

GenGenStudio synthetic data generation platform / Source: GenGenAI


iT DongA: What are GenGenAI’s future plans and goals?
Ho-jin Cho: Starting in 2025, we are aiming to increase our global recognition. Our initial focus will be on the US and European markets, given their large and active automotive and defense sectors. At CES 2025 in Las Vegas, we are planning to showcase GenGenData and GenGenVision to secure partnerships and clients. Winning the CES 2025 Innovation Award gives us confidence that we will achieve strong results.

We are also planning to launch "GenGenStudio," a synthetic data generation platform with a user-friendly interface. The platform's development is complete, and we are exploring the service models. Beyond the automotive and defense sectors, we hope to expand into manufacturing, semiconductors, CCTV, security, and healthcare.

Through continuous research and development, we will enhance synthetic data generation and acquire video generation technologies. Our ultimate goal is to become a pivotal company in the AI era that is synonymous with synthetic data.

by Man-hyuk Han (mh@itdonga.com)