Covariant develops robotic picking and placing AI systems focused on logistics and e-commerce. Its core product is the Covariant Brain, a foundation-model approach that builds a large database of real-world package sizes, shapes and materials to generalize across picking tasks. The company demonstrated its technology at ProMat and says three of its four co-founders have direct connections to OpenAI, informing its approach and analogy to generative AI. Covariant reports progress with global retailers and logistics providers and plans to apply the Covariant Brain to additional use cases across multiple sectors. Financially, the Series C extension added $75 million and brings the company’s total funding to $222 million, following an $80 million Series C announced in July 2021. The new capital will be used to further develop the Covariant Brain as a more capable foundation model and to expand deployments with retailers and logistics firms. Covariant develops the Covariant Brain, a universal AI platform that enables industrial robots to manipulate novel objects across diverse pick-and-place tasks. The company has demonstrated deployments across North America, Europe and Asia in sectors from grocery to fashion to pharmaceuticals. Existing customers include Obeta, Knapp, ABB and Bastian. Covariant currently employs just under 80 people and plans to increase headcount substantially. Financially, the startup has been actively fundraising and recently added to its capital base to support global expansion. The firm emphasizes a platform approach intended to future-proof customers' long-term modernization strategies. Covariant builds autonomy for industrial robotics, developing a universal AI intended to make robots work autonomously in real-world environments. Co-founded by UC Berkeley professor Pieter Abbeel, the three-year-old Berkeley startup came out of stealth in January. It has already deployed its technology to facilities in Europe and North America and announced a partnership with ABB in March. The company’s core product is a universal AI for robotics that can be applied across customer environments and industries. Covariant plans to use the new funding to grow headcount and to explore additional categories for its technology. The funding position reflects growing interest in robotics and automation amid the COVID-19 pandemic. Embodied Intelligence develops AI "robot brains" that can be loaded onto existing robots to enable teaching of new, complex skills. The company combines deep imitation learning, deep reinforcement learning and meta-learning to train deep neural nets from human VR demonstrations and further tune them with reinforcement learning. Users don a VR headset and guide a robot through a task; those demonstrations train the system which then continues learning until the robot can perform the task autonomously. Embodied Intelligence is targeting tasks where traditional automation breaks down, including manipulation of deformable objects (wires, fabrics, linens, apparel, fluid-bags, and food), picking parts from cluttered bins, and variable assembly work. The company plans to use the newly raised capital to develop its AI software and make robot programming accessible to non-experts. The founding team includes Pieter Abbeel, Peter Chen, Rocky Duan and Tianhao Zhang, and the company is headquartered in Emeryville, CA.

Total raised

$202M

Funding rounds

4

Latest round

Series C

Latest activity

Apr 2023