Companies are generating an unprecedented volume of video content. Ranging from decades of broadcast archives to thousands of surveillance cameras and countless hours of production footage, much of this data remains dormant on servers, unexamined and unutilized. This phenomenon is known as dark data: a vast, untapped asset that organizations automatically gather but seldom employ in a meaningful manner.
To address this challenge, Aza Kai (CEO) and Hiraku Yanagita (COO), two former Googlers who collaborated for nearly a decade at Google Japan, opted to develop their own solution. The pair co-founded InfiniMind, a Tokyo-based startup focused on building infrastructure that transforms petabytes of unviewed video and audio into structured, queryable business data.
“My co-founder, who spent a decade leading brand and data solutions at Google Japan, and I recognized this critical shift approaching while we were still at Google,” Kai explained. He added that by 2024, the technology had advanced sufficiently, and market demand had become clear enough to compel the co-founders to establish the company themselves.
Kai, whose prior experience at Google Japan spanned cloud, machine learning, ad systems, and video recommendation models, and who later led data science teams, clarified that existing solutions necessitate a compromise. Earlier approaches could identify objects in individual frames but were incapable of tracking narratives, understanding causality, or answering intricate questions about video content. For clients managing decades of broadcast archives and petabytes of footage, even basic inquiries about their content often remained unanswered.
The pivotal change, Kai observed, stemmed from the progress in vision-language models between 2021 and 2023, which is when video AI began to move beyond simple object tagging. He informed TechCrunch that while declining GPU costs and consistent annual performance improvements of approximately 15–20% over the last decade were helpful, the greater narrative was about capability; until recently, models simply couldn’t perform the required tasks.
InfiniMind recently secured $5.8 million in seed funding, with UTEC leading the round and participation from CX2, Headline Asia, Chiba Dojo, and an AI researcher at a16z Scout. The company is in the process of relocating its headquarters to the U.S., while maintaining an active office in Japan. Japan proved to be the ideal testing ground, offering robust hardware, skilled engineers, and a supportive startup ecosystem, which allowed the team to refine its technology with discerning customers before expanding globally.
Its inaugural product, TV Pulse, launched in Japan in April 2025. This AI-powered platform conducts real-time analysis of television content, enabling media and retail companies to “track product exposure, brand presence, customer sentiment, and PR impact,” according to the startup. Following pilot programs with major broadcasters and agencies, it has already acquired paying customers, including wholesalers and media organizations.
Now, InfiniMind is poised for the international market. Its flagship product, DeepFrame, a long-form video intelligence platform designed to process 200 hours of footage to precisely identify specific scenes, speakers, or events, is slated for a beta release in March, with a full launch planned for April 2026, Kai stated.
The video analysis landscape is highly fragmented. Kai noted that while companies such as TwelveLabs offer general-purpose video understanding APIs for a wide array of users, including consumers, prosumers, and enterprises, InfiniMind specifically targets enterprise use cases, encompassing monitoring, safety, security, and the extraction of deeper insights from video content.
“Our solution requires no code; clients simply provide their data, and our system processes it, delivering actionable insights,” Kai explained. “We also integrate audio, sound, and speech understanding, not just visuals. Our system can handle unlimited video length, and cost efficiency is a key differentiator. Most existing solutions prioritize accuracy or specific use cases but do not resolve cost challenges.”
The seed funding will empower the team to continue developing the DeepFrame model, expand its engineering infrastructure, recruit more engineers, and reach additional customers across both Japan and the U.S.
“This is a captivating domain, representing one of the pathways toward AGI,” Kai commented. “Achieving general video intelligence is fundamentally about comprehending reality. While industrial applications are significant, our ultimate objective is to push the boundaries of technology to foster a deeper understanding of reality and enable humans to make better decisions.”