David Z. Pan
Department of Electrical and Computer Engineering
The University of Texas at Austin
AI for Integrated Circuit Design: From Prediction to Generation and Validation
Abstract
Artificial intelligence is expanding the possibilities of integrated circuit design, from performance prediction and optimization to automated generation. This talk will provide a broad perspective on AI for chip design, followed by recent advances in analog/RF design automation spanning topology generation, device sizing, layout synthesis, and RF inverse design. We will highlight how generative AI, multimodal feedback, and agentic circuit reasoning can capture design intent and connect design stages, with examples illustrating the role of simulation, post-layout evaluation, and silicon demonstrations in validating design outcomes. We will also discuss practical opportunities and challenges in data efficiency, generalization, and robustness, and outline a path toward end-to-end IC design automation that combines AI with circuit knowledge and designer expertise.
Biography
David Z. Pan is the Silicon Labs Endowed Chair Professor at The University of Texas at Austin, and Co-Founder of DeepWeave. His research interests include design automation for digital/analog/RF ICs and emerging technologies, synergistic AI/IC co-optimizations, and design/system and technology co-optimization. He has published over 550 refereed journal/conference papers and 10 US patents. He has served in many editorial boards and conference committees, including General Chair of DAC 2027 and ICCAD 2019. He has received many awards, including SRC Technical Excellence Award, UCLA Samueli Professional Achievement in Academia Award, Nvidia Research Faculty Fellow, MediaTek Research Excellence Award, and over 20 Best Paper Awards from premier EDA/chip/AI venues. He has graduated over 60 PhD students and postdocs who are now holding key academic and industry positions. He is a Fellow of IEEE, ACM, and SPIE.
Wouter A. Serdijn
Section Bioelectronics, Delft University of Technology, Delft
Medical Delta Honorary Professor at Delft University of Technology and Erasmus Medical Center, Rotterdam
Next-Generation Integrated Circuits for Neural Implants: artifact-resilient recording, power-efficient stimulation and analog edge-AI
Abstract
Understanding and treating brain disorders increasingly requires modulation of distributed brain networks rather than individual neural targets. The NeuroDots concept envisions a coordinated ensemble of miniature implantable devices that can sense and stimulate multiple interacting brain regions simultaneously [1]. Realizing such a vision demands highly integrated circuits and systems that combine ultra-low-power sensing, wireless communication, adaptive neuromodulation, and intelligent closed-loop control.
Beyond recording neural activity, these systems must acquire multimodal physiological information and use it to infer symptom-related network states. Self-learning algorithms operating at the edge will need to recognize patient-specific patterns of brain activity and connectivity and dynamically select stimulation strategies that alleviate symptoms by reshaping dysfunctional network interactions.
This talk explores how NeuroDots could enable next-generation brain-network modulation and discusses the integrated circuit and system innovations required to make them a reality. Topics include the network organization of the brain, energy-efficient neural stimulation, multi-channel neural recording using code-division multiplexing and compressed sensing, and emerging analog edge-AI paradigms for implantable neurotechnology. Together, these developments point toward a future in which large-scale networks of intelligent implantable devices can interact with the brain at the level at which neurological and psychiatric disorders emerge: the brain network itself.
[1] De Ridder, Dirk et al. “NeuroDots: From Single-Target to Brain-Network Modulation: Why and What Is Needed?” Neuromodulation: journal of the International Neuromodulation Society vol. 27,4 (2024): 711-729. doi:10.1016/j.neurom.2024.01.003
Biography
Wouter A. Serdijn is a full professor in bioelectronics at Delft University of Technology, where he chairs the Section Bioelectronics, and a Medical-Delta honorary professor at both Delft University of Technology and the Erasmus Medical Center, Rotterdam.
His research interests include integrated biomedical circuits and systems for wearable, injectable and implantable medical devices, such as cardiac pacemakers, cochlear implants, neurostimulators, bioelectronic medicine and electroceuticals.
Wouter A. Serdijn is an IEEE Fellow, an IEEE Distinguished Lecturer and a mentor of the IEEE. In 2016, he received the IEEE Circuits and Systems Meritorious Service Award.
More information can be found at: https://bioelectronics.tudelft.nl/~wout/.