TechCrunch Industry News Ex-Googlers are building infrastructure to help companies understand their video data
Feb 9, 2026
They explore how companies sit on vast troves of unused video data and the challenge of turning it into searchable business intelligence. The founders' Google Japan roots and a move to the U.S. are highlighted. They discuss advances in vision-language models, no-code multimodal tools, and products that analyze TV and long-form video in real time.
AI Snips
Chapters
Transcript
Episode notes
Video Archives Are Untapped Goldmines
- Companies accumulate vast amounts of video that remain unused as 'dark data' on servers.
- InfiniMind aims to convert these archives into structured, queryable business intelligence.
Founders' Google Roots And Japan Testbed
- Azakai and Hiraku Yanagita left Google Japan to found InfiniMind after seeing the market inflection.
- They used Japan as a demanding testbed before relocating HQ to the U.S. after raising seed funding.
Vision-Language Advances Enabled Video AI
- Progress in vision-language models between 2021–2023 enabled video AI to move beyond frame-level tagging.
- Falling GPU costs and steady accuracy gains made deeper video understanding practical.