On-device AI software startup ZETIC.ai, led by CEO Yeonseok Kim, released the beta version of their integrated on-device AI solution, ZETIC.MLange, on the 29th.

Under the brand mission "Bring on-device AI to everyone," ZETIC.ai is developing on-device AI solutions that enable startups offering AI services to internalize AI computations typically handled by servers. In May, ZETIC.ai secured seed funding from Korea Investment Accelerator and The Ventures.

ZETIC.MLange is an automated solution that transforms existing AI models (such as STT, TTS, Image Segmentation, and Generative AI) into on-target AI model libraries, integrating them into on-device AI. With ZETIC.MLange, users can easily, quickly, and conveniently create on-device AI applications using their existing AI models.

One of the significant advantages of ZETIC.MLange is its economic efficiency, as it allows the internalization of costs associated with server calls into the device, saving expenses. ZETIC.MLange supports all operating systems, processors, and devices, making it easy to implement on-device AI applications without complex engineering processes.

Previously, developing on-device AI applications required customers with AI models to use NPUs from on-device AI hardware companies like Furiosa and Rebellion or model optimization tools from on-device AI software companies like Nota AI and SqueezeBits. However, this process often took several months to years, representing a major pain point in the industry.

ZETIC.MLange is designed with a developer-friendly UX, prioritizing ease of use. It eliminates the complicated processes required for on-device AI conversion, allowing users to generate a Melange Key from an AI model and obtain an on-device AI application with a single command. This means that customers can create on-device AI applications within 24 hours without additional engineering work, as long as they have an AI model.

Current on-device AI hardware companies are limited to supporting only devices integrated with the NPUs they have developed. On-device AI software companies also face challenges with insufficient NPU support or requiring customers to handle the implementation themselves.

ZETIC.MLange supports various NPUs from different SoC companies and a range of operating systems, enabling on-device AI implementation regardless of the device. Currently, ZETIC.MLange supports Android, iOS, Linux OS, and NPUs from Qualcomm, MediaTek, and Apple.

Traditionally, AI service companies had to rely on server calls for service operations. For example, OpenAI incurs server costs amounting to tens of billions of won per month. The cumulative AI infrastructure costs have been a significant burden for AI service companies.

To address these pain points, ZETIC.MLange eliminates the need for server costs through on-device technology. Customers only need to pay a one-time setup fee and license costs, allowing for more economical AI service operations.

ZETIC.ai has now released the beta version of ZETIC.MLange. Anyone can use ZETIC.MLange through the Product page on the ZETIC.ai website. Currently, ZETIC.ai offers two pricing plans: Trial and Enterprise. They also plan to introduce new pricing plans that will utilize Melange by the end of the year.