Covariant :: Reliable, Flexible & Scalable AI Robotics

Covariant is a leading AI robotics company headquartered in Emeryville, California. Founded in 2017 (originally as Embodied Intelligence) by world-renowned AI researchers including Pieter Abbeel, the company develops Robotics Foundation Models that power intelligent, general-purpose robotic systems for warehouse automation. Covariant’s mission is to deliver reliable, flexible, and scalable AI robotics that address labor shortages and operational complexity in modern fulfillment centers. Its flagship technology, the Covariant Brain, is a universal AI platform trained on the world’s largest multimodal robotics dataset from real warehouses. This enables robots to “see, think, and act” with human-like adaptability, handling virtually any SKU or item on Day One without extensive reprogramming.

Covariant focuses on practical, production-grade automation rather than research prototypes. Its AI-powered robots are deployed in partnership with leading warehouse integrators (such as KNAPP) and major fulfillment providers like Radial, automating high-variability tasks that were previously manual and labor-intensive.

Key Robot-Based Solutions:

Covariant’s Covariant Brain powers multiple robotic use cases across the warehouse:

• Robotic Goods-to-Person Picking: Robots pick from AutoStore, shuttles, AMRs, and other AS/RS systems, turning traditional goods-to-person setups into fully autonomous goods-to-robot systems. This increases efficiency and closes labor gaps.
• Robotic Induction: Automates the manual process of inducting parcels, polybagged apparel, pharmaceuticals, and other items onto unit sorters, pocket sorters, AMRs, and auto-baggers. It handles chaotic, tightly packed, or edge-case items with high reliability.
• Robotic Depalletization: Automates the unloading of mixed-SKU pallets onto conveyors, ensuring timely replenishment of storage and pick zones while reducing physical strain on workers.
• Robotic Kitting: Supports co-packing, meal kitting, and subscription box assembly with precise, adaptive manipulation.

The Covariant Brain (powered by RFM-1, the Robotics Foundation Model) learns continuously through fleet learning, improving performance across all connected robots. It excels at handling extreme variability in shape, size, packaging, and orientation. Covariant has demonstrated superior performance in competitions (e.g., winning the ABB Order Picking Competition) and real deployments, delivering consistent, high-quality results that outperform manual labor in speed, accuracy, and reliability. The company continues to advance its foundation model, expanding capabilities and partnerships to power the next generation of warehouse automation.