
Web data infrastructure for AI training
wyndlabs.ai (opens in a new tab)Wynd Labs builds infrastructure that delivers web data at volume to the companies training the world's most powerful AI models. It powers and supports Grass, a bandwidth-sharing network that lets it operate a distributed crawler with unusual access to high-quality public web data globally, and has built pipelines for ingesting, segmenting and annotating billions of videos, transcripts and audio files for frontier lab datasets. The distributed approach is the substantive part: crawling the open web at that scale from central infrastructure runs into limits a bandwidth-sharing network does not.