The competitiveness of on-device AI comes from AI optimization technology. On-device AI has become a core component of modern smart technologies. Notable examples include Apple Intelligence, recently introduced by Apple, as well as applications in autonomous vehicles and edge systems within smart factories. While the rise of generative AI has increased the importance of performing AI computations directly on devices, on-device AI faces significant limitations due to constrained hardware resources such as computing power, memory capacity, and battery life.
[Fig] Characteristics and Advantages of On-Device AI
Source: Mirae Asset Global Investments
To realize on-device AI, AI optimization technology is essential. AI optimization can generally be categorized into three major areas: AI model compression, hardware optimization, and algorithm optimization.
AI Model Compression Technology simplifies model architectures by retaining only essential components while minimizing unnecessary parameters and structures.
Hardware-Based AI Optimization involves implementing AI-specific hardware suited for on-device environments, such as integrating Neural Processing Units (NPUs) or connecting multiple GPUs in parallel to improve computational performance.
Algorithm Optimization Technology improves numerical stability by minimizing computational errors and enhances efficiency through optimized memory access patterns.
These optimization technologies are being actively adopted by various on-device AI companies, including Kakao Pay, Hyperconnect, and QANDA.
[Fig] Overview of Model Compression and Algorithm Optimization Technologies for AI Optimization
Korean startups are actively developing AI optimization technologies for on-device AI environments. One notable company is OptAI Co., Ltd. (CEO: Jaeho Lee). OptAI develops lightweight AI models tailored to specific device environments and has created automated deployment technologies for on-device AI applications. Through these capabilities, the company has secured major clients such as LG Electronics and Hyundai Motor Company. OptAI was recently selected for the Deep Tech TIPS (Super Gap Startup 1000+ Project) program. With funding support of up to KRW 1.5 billion over three years, the company plans to advance its existing optimization technologies and develop solutions capable of optimizing large-scale foundation AI models.
[Fig] OptAI’s AI Optimization Technology
Another AI solution provider, Nota Inc., is also actively developing AI optimization technologies for on-device AI. Nota addresses compatibility issues arising from unsupported nodes and hardware environments by applying optimized AI models tailored to specific devices. This enables highly optimized AI execution environments on edge devices.
In addition, Algorithm Labs Co., Ltd. (CEO: Jinho Son) provides on-demand AI solutions customized to industry-specific requirements based on its proprietary AI platform. The company has developed optimization technologies utilizing artificial neural network-based fusion network architectures, enabling the enhancement and optimization of AI models. Through these technologies, existing AI models can achieve improved problem-solving performance and greater operational efficiency.
As demonstrated by these examples, companies are actively investing in optimization technologies that enable practical deployment of on-device AI.
[Fig] Nota’s On-Device AI Optimization Solution “Netspresso”
The Korean government is also actively supporting the development of on-device AI technologies that rely on AI optimization. The Ministry of SMEs and Startups (MSS) launched the “On-Device AI Super Gap Challenge” to identify and foster promising startups in this field. Participating companies receive technical validation (PoC) and technical support from global industry leaders including LG Electronics (devices), Intel (chipsets), and Microsoft (software), while the government provides collaboration and development funding. Such support programs are expected to expand further as the government seeks to accelerate the advancement of on-device AI technologies.
#AIOptimization #OnDeviceAI #ModelCompression #AlgorithmOptimization #GenerativeAI #AppleIntelligence #KakaoPay #Hyperconnect #QANDA #OptAI #Nota #AlgorithmLabs
The competitiveness of on-device AI comes from AI optimization technology. On-device AI has become a core component of modern smart technologies. Notable examples include Apple Intelligence, recently introduced by Apple, as well as applications in autonomous vehicles and edge systems within smart factories. While the rise of generative AI has increased the importance of performing AI computations directly on devices, on-device AI faces significant limitations due to constrained hardware resources such as computing power, memory capacity, and battery life.
[Fig] Characteristics and Advantages of On-Device AI
Source: Mirae Asset Global Investments
To realize on-device AI, AI optimization technology is essential. AI optimization can generally be categorized into three major areas: AI model compression, hardware optimization, and algorithm optimization.
AI Model Compression Technology simplifies model architectures by retaining only essential components while minimizing unnecessary parameters and structures.
Hardware-Based AI Optimization involves implementing AI-specific hardware suited for on-device environments, such as integrating Neural Processing Units (NPUs) or connecting multiple GPUs in parallel to improve computational performance.
Algorithm Optimization Technology improves numerical stability by minimizing computational errors and enhances efficiency through optimized memory access patterns.
These optimization technologies are being actively adopted by various on-device AI companies, including Kakao Pay, Hyperconnect, and QANDA.
[Fig] Overview of Model Compression and Algorithm Optimization Technologies for AI Optimization
Korean startups are actively developing AI optimization technologies for on-device AI environments. One notable company is OptAI Co., Ltd. (CEO: Jaeho Lee). OptAI develops lightweight AI models tailored to specific device environments and has created automated deployment technologies for on-device AI applications. Through these capabilities, the company has secured major clients such as LG Electronics and Hyundai Motor Company. OptAI was recently selected for the Deep Tech TIPS (Super Gap Startup 1000+ Project) program. With funding support of up to KRW 1.5 billion over three years, the company plans to advance its existing optimization technologies and develop solutions capable of optimizing large-scale foundation AI models.
[Fig] OptAI’s AI Optimization Technology
Another AI solution provider, Nota Inc., is also actively developing AI optimization technologies for on-device AI. Nota addresses compatibility issues arising from unsupported nodes and hardware environments by applying optimized AI models tailored to specific devices. This enables highly optimized AI execution environments on edge devices.
In addition, Algorithm Labs Co., Ltd. (CEO: Jinho Son) provides on-demand AI solutions customized to industry-specific requirements based on its proprietary AI platform. The company has developed optimization technologies utilizing artificial neural network-based fusion network architectures, enabling the enhancement and optimization of AI models. Through these technologies, existing AI models can achieve improved problem-solving performance and greater operational efficiency.
As demonstrated by these examples, companies are actively investing in optimization technologies that enable practical deployment of on-device AI.
[Fig] Nota’s On-Device AI Optimization Solution “Netspresso”
The Korean government is also actively supporting the development of on-device AI technologies that rely on AI optimization. The Ministry of SMEs and Startups (MSS) launched the “On-Device AI Super Gap Challenge” to identify and foster promising startups in this field. Participating companies receive technical validation (PoC) and technical support from global industry leaders including LG Electronics (devices), Intel (chipsets), and Microsoft (software), while the government provides collaboration and development funding. Such support programs are expected to expand further as the government seeks to accelerate the advancement of on-device AI technologies.
#AIOptimization #OnDeviceAI #ModelCompression #AlgorithmOptimization #GenerativeAI #AppleIntelligence #KakaoPay #Hyperconnect #QANDA #OptAI #Nota #AlgorithmLabs