Hiemdex Selected for South Korea’s Ministry of SMEs and Startups ‘Stepping Stone R&D’ Program to Advance ‘Zero-Upload’ AI Video Search and Editing

SEOUL, South Korea — Hiemdex, a pioneering AI startup specializing in "zero-upload" video search, editing, and utilization, announced today that it has been officially selected for the 2026 Startup Growth Technological Development (Stepping Stone) R&D Program, hosted by the South Korean Ministry of SMEs and Startups (MSS).

Through this government-backed initiative, Hiemdex will focus on advancing its core technologies: inferring user search intent within massive video archives using vision-language models, and establishing advanced methods for measuring search accuracy.

Solving the Limitation of Keyword Searches with Natural Language

The core objective of this R&D project is to enable technology that can pinpoint exact video segments using long, descriptive sentences. For instance, a user could search for a highly specific moment, such as: "The exact moment the show host lifts the product and the price caption first appears on the screen."

While most conventional video search systems are optimized for short keyword matching—often failing when users describe scenes in detail—Hiemdex aims to solve this pain point. The company's technology directly aligns and maps lengthy, descriptive natural language queries to corresponding moments within the video footage.

Privacy-First, Vector-Native Architecture

Hiemdex provides natural language video search, automated short-form video generation, and compliance processing within a unified platform. Distinctively, it operates on a "zero-upload" (vector-native) on-premise architecture, meaning raw, high-volume video files do not need to be transmitted to external servers, ensuring maximum data privacy.

This technology holds immense potential across industries managing vast archives of unstructured video data, including broadcasting, e-commerce, legal services, CCTV security, and robotics.

To accelerate this research, Hiemdex is partnering with the Perceptual AI Lab at Chung-Ang University, led by Professor Chanho Eom. Professor Eom’s lab brings deep expertise in computer vision, specializing in video search and person re-identification (Re-ID).

"This selection is a testament to the government’s validation of Hiemdex’s technological prowess and market potential," said Heejo Kang, Co-founder of Hiemdex. "By enabling users to find exact scenes using natural language without ever uploading their videos externally, we aim to deliver a highly differentiated solution for the broadcasting and e-commerce sectors, where data security is paramount."

With this government recognition and funding, Hiemdex is poised to accelerate its expansion into the global media, live commerce, and entertainment markets.

하임덱스, 중기부 ‘디딤돌 R&D’ 선정…‘업로드 없는’ 영상 검색·편집 AI 고도화 나서
업로드 없는 영상 검색·편집·활용 AI 전문기업 하임덱스(대표 이장원)가 중소벤처기업부가 주관하는 ‘2026년도 창업성장기술개발사업(디딤돌) R&D’ 과제에 최종 선정됐다고 밝혔다.하임덱스는 이번 과제를 통해 대용량 동영상 아카이브에서 비전-언어 기반으로 사용자의 검색 의도를 추론하는 기술과 검색 정확도 측정 방법을 연구한다.이번 과제의 핵심은 “쇼호스트가 제품을 들어 보이다가 가격 자막이 처음 뜨는 순간”처럼 길고 서술적인 문장만으로 원하는 영상 구간을 찾아내는 기술이다. 대다수 영상 검색 시스템이 짧은 키워드 매칭에 최적화돼 있