The ITW (In-The-Wild) Dataset from Noitom Robotics is a human-centric, life-scale, diverse, open-world, multimodal dataset designed to advance humanoid robotics and embodied intelligence. It includes sparse-body motion capture sensors and egocentric RGB-D visual data captured in unconstrained real-world environments. By incorporating high-variance environmental noise and long-tail edge cases, it facilitates the generalization of laboratory-optimized algorithms toward industrial deployment.