Moreh
Moreh, founded in 2020, develops the MoAI platform, a full-stack infrastructure software for the LLM era, supporting over 2,000 GPUs and powering large-scale AI models.
Company Overview
Moreh was founded in 2020 and specializes in developing the MoAI platform, a comprehensive infrastructure software designed for the Large Language Model (LLM) era. With headquarters in Santa Clara, CA, and additional offices in Seoul and Hanoi, Moreh is positioned both in the U.S. and Asia.
MoAI Platform
The MoAI platform is a full-stack infrastructure software from PyTorch to GPUs, tailored for the Large Language Model (LLM) era. It currently supports AMD ROCm and will expand to include other accelerators. The platform is used for programming and scaling models with multi-billion or even multi-trillion parameters. One of its key features includes a virtualization layer for GPUs that ensures higher utilization and failover capabilities.
Funding and Financials
Moreh has raised a total of $30 million in funding, with $22 million coming from its recent Series B round. This capital injection supports its continuous innovation and expansion of the MoAI platform.
Global Operations
Moreh's global operations span headquarters in Santa Clara, CA, and additional offices in Seoul, South Korea, and Hanoi, Vietnam. This geographical diversity allows the company to serve a wide range of markets with its advanced LLM infrastructure solutions.
Commercial Applications
The MoAI platform powers over 2,000 GPUs for commercial Infrastructure as a Service (IaaS) and Platform as a Service (PaaS) clouds. It is also utilized in on-premise LLM training systems. One notable deployment is for KT, where the platform was used to train large AI models, proving effective in overcoming public cloud service challenges.
Significant Achievements
Moreh trained the largest-ever Korean LLM with 221 billion parameters using a cluster of 1,200 AMD MI250 GPUs. This significant achievement underscores the scalability and effectiveness of the MoAI platform in handling large-scale AI model training.