A San Francisco startup and a major Korean logistics company are taking a different approach to warehouse automation: they're teaching robots to feel and manipulate objects before teaching them to understand language. RLWRLD and CJ Logistics announced a strategic partnership to develop and deploy robotics foundation models (RFMs) specifically designed for logistics tasks, moving beyond the lab to real-world warehouse operations.

What Makes This Partnership Different From Other Robot Development?

Most AI companies building robots start with language models and visual understanding, then layer on physical capabilities. RLWRLD is flipping that script. The company's philosophy centers on what it calls "dexterity before intelligence," meaning robots need to master complex hand movements and tactile feedback before they can reason about what they're doing.

"Most approaches in this market focus on intelligence first, making language models lead to visual understanding and then action. We take a 'dexterity before intelligence' approach. We met with 200 industry leaders, and most said that a high level of dexterity is needed to do warehouse and factory tasks," said Junghee Ryu, CEO of RLWRLD.

Junghee Ryu, CEO of RLWRLD

This insight came from conversations with 200 industry leaders who emphasized that warehouse and factory work demands precision manipulation. Ryu noted that robots need tactile, force-torque, and motion information, not just visual data or language understanding.

How Are They Building Smarter Warehouse Robots?

• Five-Fingered Hands: RLWRLD chose five-fingered hands because many real-world logistics tasks require the dexterity that human-like hands provide, rather than simpler gripper designs.
• High-Quality Training Data: The company combines open-source data with proprietary information, using simulation to generate synthetic training data extracted from videos, avoiding the lower-quality data that comes from teleoperators in some "data factories."
• Real-World Deployment: Rather than validating technology only in labs, RLWRLD and CJ Logistics will deploy systems in actual logistics operations, collect data from those real environments, and feed that information back into continuously improving AI models.
• Foundation Model Focus: RLWRLD's RLDX-1 foundation model, released in May, uses visual and sensor inputs to give humanoid robots environmental understanding and the ability to execute complex tasks appropriate to warehouse conditions.

The partnership combines RLWRLD's foundation model and software capabilities with CJ Logistics' operational sites, infrastructure, and commercialization expertise. Together, they will identify specific logistics processes for initial deployment and conduct proof-of-concept projects to validate the technology.

What's the Long-Term Vision for This Technology?

The companies' ultimate goal is ambitious: commercializing "logistics as a service" (LaaS) offerings by 2029. In this model, businesses would use logistics capabilities as a service rather than building and owning their own logistics centers, transportation networks, and automation infrastructure.

"For physical AI to translate into tangible productivity gains in industry, it is essential to go beyond validating technology in the lab, accumulate data from real-world operations, and feed that data back into continuously improving AI intelligence," explained Junghee Ryu.

Junghee Ryu, CEO of RLWRLD

LaaS would encompass a wide range of functions, from order processing and warehousing to picking, packing, delivery, returns, and data analytics. This shift represents a fundamental change in how companies approach logistics infrastructure.

RLWRLD, founded in 2024, has already raised $41 million in total funding, including a $26 million seed round in February. The company is currently engaged in pilots with multiple partners and has announced collaborations with NVIDIA and other industry players. Ryu noted that the company is preparing for commercial use through partnerships with industry leaders who understand that hardware and software providers must collaborate effectively.

The partnership also positions both companies to expand beyond South Korea's distinctive logistics environment. RLWRLD's global partner network combined with CJ Logistics' commercialization expertise will enable joint go-to-market initiatives and strategic alliances in overseas markets, particularly in the United States, Japan, and Taiwan.