조강원 대표

모레

발표제목: Infrastructure Software for Hyperscale AI / 발표초록: With the growing interest in LLMs and multimodal AI, many companies and people are struggling to construct and utilize large-scale AI infrastructure with thousands or more accelerators. The most difficult challenges in AI technology are now a variety of infrastructure-level problems such as parallelization, cluster scalability, performance portability, orchestration of heterogeneous accelerators, and failover. Moreh is a startup that believes many infrastructure-level challenges are mainly due to the limitations of the legacy AI software stacks, specifically deep learning frameworks and parallel computing platforms. From this insight, it has been developing the MoAI platform, a set of fully integrated software components from deep learning primitive libraries to application-level APIs. It provides 100% PyTorch/TensorFlow compatibility and ensures the same user experience as traditional AI software stacks. It supports single device abstraction and application-level virtualization with the help of the just-in-time graph compiler, to bridge AI applications and underlying accelerators in a more efficient, scalable, and flexible way.