Markov builds datasets and environments for training computer-use AI, including workflow data, synchronized interaction signals, verifiable outcomes, CAD environments, browser-use data, coding data, design data, spreadsheet data, and gaming recordings.