[Abstract]

Artificial intelligence offers new opportunities to automate complex hardware design tasks that traditionally require substantial human expertise and iterative engineering effort. In this talk, I will present two examples of AI for hardware design automation spanning different stages of the semiconductor design lifecycle. First, I will introduce an AI-based framework for NAND flash post-fabrication optimization, where a large number of device parameters must be calibrated to achieve target performance and reliability. The proposed approach combines global exploration and local refinement to robustly identify high-quality configurations even from poor initial solutions. I will then present AnalogToBi, a framework for device-level analog circuit topology generation. AnalogToBi combines a bipartite graph representation with grammar-guided decoding to generate electrically valid and structurally novel circuit topologies without human intervention during training. Together, these studies demonstrate how AI can automate both pre-fabrication design and post-fabrication optimization by incorporating hardware-specific structures and constraints into the learning and search process.

[Biography]

Yulhwa Kim received the B.S. and Ph.D. degrees in Convergence IT Engineering from Pohang University of Science and Technology, South Korea, in 2016 and 2022, respectively. From 2022 to 2024, she held a postdoctoral position in the Inter-University Semiconductor Research Center (ISRC) at Seoul National University, South Korea. She is currently an assistant professor in the Department of Semiconductor Systems Engineering at Sungkyunkwan University, South Korea. Her recent research focuses on AI for hardware design automation. Her research also includes hardware-software co-design for efficient AI systems through neural network compression and deep learning accelerator design.