[Startup CEOs of Hannam University’s Entrepreneurship-Focused University Program] techniflows, Advancing Secure Data Utilization with On-Device AI and Lightweight MLOps Platforms

Donghoon Kim CEO of techniflows (2025 Selected General Companies within the Region)

-Flagship products, ‘VideoMasker’ (Video De-identification) and ‘miniOps’ (Lightweight MLOps)

-High accessibility and convenience through an automated, web-based mosaic solution requiring no installation

techniflows is a specialized AI enterprise developing on-device video and image de-identification solutions and lightweight MLOps platforms that simplify AI development even in small-scale GPU environments. The company was founded in May 2025 by CEO Donghoon Kim (30).

One of the core products, VideoMasker, is an on-device de-identification service that automatically mosaics sensitive information—such as faces and license plates—directly in a web browser without requiring any software installation.

Since all processing is performed locally on the user's device rather than uploading videos to a server, the risk of personal data leakage is eliminated. Furthermore, by utilizing face recognition, the system can accurately identify the same person across different scenes, allowing users to selectively de-identify only specific individuals throughout an entire video.

VideoMasker holds three primary competitive advantages. Accessibility & Convenience, An automated mosaic service that runs directly in a web browser with zero installation.

Security & Regulatory Compliance, Its on-device architecture ensures all video processing stays within the user's device, preventing external data transfer.

Selective Tracking, Using advanced facial recognition, it consistently tracks specific targets even through frequent cut transitions, ensuring stable and selective de-identification. These factors distinguish VideoMasker from manual editing, outsourcing, or server-based solutions.

“While developing VideoMasker, I realized that AI services require constant management, including repeated model updates and functional improvements”, said CEO Kim. “I recognized that a variety of systems and automation are necessary to perform these tasks efficiently. The technology we created to solve this directly is miniOps.”

miniOps is a lightweight MLOps platform designed to easily automate AI development and operations in small-scale GPU environments. It provides a structured workflow that automates the entire lifecycle—including data preparation, training execution, model management, and deployment—without the need for complex infrastructure. The AI model management system within VideoMasker itself is powered by miniOps, and techniflows has expanded this technology into a form that any team can use to build and maintain AI services.

The strength of miniOps lies in its ability to automate the entire AI development pipeline. It allows organizations to manage data preparation and deployment processes—tasks that typically require significant time and specialized manpower—without additional experts. Even with an environment of only 1 to 4 GPUs, a team can establish an enterprise-level AI development and operation system. Since the technology was validated in the real-world environment of VideoMasker, its reliability is well-established, allowing researchers to focus on actual AI research rather than complex infrastructure building.

Regarding future plans, CEO Kim stated, “While AI technology is advancing rapidly, it remains concentrated in the hands of a few corporations and experts. We want to change this. We plan to continue making AI models lighter and safer, expanding our technology into a form that anyone can use without burden through on-device solutions.”

Date of Establishment: May 2025
Core Business: Development of on-device AI-based video/image de-identification solutions; Provision of lightweight MLOps (miniOps) platforms
Key Achievements: Selected for the 2025 Hannam University Startup Hub Project, Participated in TECHFEST Vietnam 2025 (Daejeon Startup Springboard Program), Finalized for the IP Step-Stone Program (Patent application completed), Selected for the Regional Enterprise Growth Program (Chungnam National University & Korea Gas Technology Corp), Grand Prize at the DDU Startup Idea Competition

reporter jinho lee
jinho2323@hankyung.com