Claru

Claru builds training data for physical AI: purpose-built egocentric video and robotics datasets captured in kitchens, workshops, warehouses, agricultural settings, driving environments, and other real-world tasks, enriched with depth maps, pose estimation, segmentation masks, and structured metadata for robotics, embodied-AI, and world-model companies. It describes itself as the only training-data company 100% focused on physical AI. Claru operates as the training-data platform of Reka AI, a foundation-model lab founded in 2022 by Dani Yogatama (CEO), Yi Tay (Chief Scientist), Cyprien de Masson d'Autume (CTO), Qi Liu, and Mikel Artetxe, headquartered in San Francisco, which has raised $168 million and reached a valuation of roughly $1 billion from investors including NVIDIA, DST Global, and Snowflake. Claru itself launched in 2025 and is backed by the same investor group, including General Catalyst, Khosla Ventures, Bessemer Venture Partners, and NVIDIA. In June 2026, Reka added a research and engineering team from Moonvalley, including former Google DeepMind researchers Mateusz Malinowski and Mikolaj Binkowski, who had worked on Google's Veo video models, to accelerate physical-AI model and infrastructure development alongside Claru's data layer. Claru reports a network of more than 10,000 trained collectors across more than 100 cities on 6 continents, more than 500,000 egocentric videos, more than 386,000 annotated clips, more than 4 million human annotations, and more than 100 active datasets, delivered to 5 or more frontier AI lab partnerships.

Claru has two different kinds of work, and it matters which one you're applying for. Its core team is a small engineering and research group, described as drawing ex-FAANG talent, working on the data-enrichment pipeline (depth maps, pose estimation, segmentation, optical flow) and the platform connecting collectors to frontier AI labs; exact headcount is not publicly disclosed. The much larger group is its network of 10,000+ trained collectors, people who wear cameras to capture egocentric video during real household, workshop, warehouse, agricultural, or driving tasks, plus expert human annotators who add action-boundary labels, object affordances, grasp types, and natural-language instruction annotations. Collector and annotator work is contractor-based, not salaried employment. Confirmed posted roles include household-chores video capture in the United States (paying $15/hour for accepted footage, via biweekly direct deposit) and in Latin America (paying $6/hour, via Wise, Payoneer, or local transfer). Application flow for collector roles lives at claru.ai/jobs; core team roles would run through Reka AI's own hiring process. Check each posting for exact task, location eligibility, and pay.