Artificial Intelligence
AI models are at the center of computing architecture changes, and we are closely analyzing AI model structures, training methods, and the evolution of software platforms that support them. We quantify the memory and storage requirements necessary for these changes, and use them as a basis to suggest the direction of next-generation memory and storage solutions. We are also conducting research on the optimization potential of AI models themselves or training or inference platforms.

Want to learn more about how we can create new values for semiconductor industry? Please contact us!
We are in this together! Contact Us![[Summary] MoSKA: Mixture of Shared KV Attention for Efficient Long-Sequence LLM Inference](https://mis-prod-koce-research-user-cdn-01-blob-ep.azureedge.net/web/blog/20260324/thumb_wxj3sRn5.20260324141825249.jpg)
![[Summary] Proactive Embedding on Cold Data for Deep Learning Recommendation Model Training](https://mis-prod-koce-research-user-cdn-01-blob-ep.azureedge.net/web/blog/20250917/thumb_XlRJ7RRZ.20250917093833417.png)