Shofo is building the world is largest video library by combining public videos from the open web with private videos aggregated from thousands of different sources. The result is a single, continuously updating index of billions of clips that AI labs can query for highly specific training data. Where teams once waited months to source niche footage, Shofo can deliver curated, labeled datasets in days.
Founded in 2025 by Bryan Hong, Andre Braga, Braiden Dishman, and Alexzendor Misra, the team previously built Correkt, a multimodal AI search engine that reached 40,000 users before pivoting to focus on the data layer powering modern video models. The founders bring experience from MIT, AWS, and Berkeley, and now operate out of San Francisco as part of Y Combinator W26 batch with four full-time employees.
The platform deploys agents to find, clean, and label custom subsets on demand, returning datasets such as 100,000 hours of cooking footage with reasoning annotations or other narrowly scoped collections that would otherwise require an entire data operations team. By acting as a Common Crawl for video, Shofo aims to become the default substrate for video foundation models, world models, and embodied AI systems being built across frontier labs.