Junsoo Kim

Title: Junsoo Kim: Advancing Efficient AI Hardware and Inference

Introduction

Junsoo Kim is an inventor based in Hwaseong, South Korea. His patent work addresses hardware mapping for generative artificial intelligence models and multi-device inference for large language models. He is associated with Hyperaccel Co., Ltd. and has 4 patents.

Latest Patents

One of Kim’s latest patents, “Method and System for Efficient Hardware Mapping of Generative Giant Artificial Intelligence Model,” describes a process for mapping model software to hardware. The method uses sequential source-code-level, instruction-level, and register-transfer-level simulations to support efficient hardware mapping.

Another patent, “Method and System for Performing Multi-Device Based Inference for Large Language Model,” concerns distributing a large language model across multiple devices using intra-layer parallelism. The devices synchronize by sharing matrix-multiplication sub-results while computation is underway.

Career Highlights

Kim’s patent portfolio reflects work on improving the implementation and execution of large AI models. His inventions focus on simulation-based hardware mapping and coordinated inference across multiple devices.

Collaborations

Kim’s listed coworkers include Seongmin Hong and Gyubin Choi. His work is associated with Hyperaccel Co., Ltd.

Conclusion

Junsoo Kim’s inventions address key engineering challenges in deploying generative AI and large language models, with an emphasis on efficient hardware mapping and multi-device inference.