Ex-Googlers Build Tools to Unlock Video Data Insights

Companies today are sitting on a massive, untapped asset: video. From decades of broadcast archives and retail security feeds to endless hours of production footage, this content typically remains stored and forgotten, representing a form of dark data that holds immense potential value. A new venture founded by former Google engineers aims to transform this dormant resource into a powerful engine for business intelligence.

The startup, InfiniMind, was co-founded in Tokyo by Aza Kai and Hiraku Yanagita, who collaborated for nearly ten years at Google Japan. They identified a critical gap in the market. While businesses accumulate petabytes of video, existing analytical tools were insufficient. Earlier technologies could identify objects in single frames but fell short of tracking narratives, understanding cause-and-effect relationships, or answering complex questions about the content. For clients with decades of broadcast archives, even basic questions about their content often went unanswered.

A technological shift between 2021 and 2023 changed the game. Advances in vision-language models moved video artificial intelligence beyond simple object tagging. According to Kai, while falling GPU costs and steady performance improvements helped, the core breakthrough was in capability, previous models simply couldn’t perform the required tasks. This inflection point convinced the duo to build their own infrastructure to convert vast stores of unviewed video and audio into structured, queryable data.

InfiniMind recently closed a $5.8 million seed funding round led by UTEC, with participation from CX2, Headline Asia, Chiba Dojo, and an AI researcher affiliated with a16z Scout. The company is now relocating its headquarters to the United States while maintaining its Japanese operations. Japan served as an ideal initial market, offering robust hardware, engineering talent, and a supportive ecosystem that allowed the team to refine its technology with demanding local customers before a global expansion.

Its first commercial product, TV Pulse, launched in Japan in April 2025. This AI platform analyzes television broadcasts in real time, enabling media and retail firms to monitor product exposure, track brand presence, gauge customer sentiment, and measure public relations impact. Following successful pilots with major broadcasters and agencies, the startup has secured paying customers, including wholesalers and media companies.

The company’s upcoming flagship offering is DeepFrame, a long-form video intelligence platform. Designed for enterprise use cases like monitoring, safety, and security, it can process up to 200 hours of footage to locate specific scenes, speakers, or events. A beta release is scheduled for March, with a full launch planned for April 2026.

Kai emphasizes that the video analysis sector is currently fragmented. While some companies offer general-purpose video understanding APIs for a wide audience, InfiniMind concentrates specifically on deep enterprise insights. A key differentiator is a no-code solution where clients provide their data, and the system handles processing to deliver actionable intelligence. The technology also integrates audio and speech understanding alongside visual analysis, supports unlimited video length, and prioritizes cost efficiency, an area where many existing solutions fall short.

The newly acquired seed funding will fuel further development of the DeepFrame model, expand engineering infrastructure, support new hires, and drive customer acquisition in both Japan and the U.S. For the founders, this work represents more than just an industrial application; it’s a step toward broader artificial intelligence goals. Understanding general video intelligence, Kai suggests, is fundamentally about understanding reality. The ultimate mission is to push technological boundaries to enhance human decision-making.

(Source: TechCrunch)