Cunxi Yu (University of Maryland, College Park)
When AI Agents Meet Chip Design: The Power of Domain Knowledge
Abstract
Autonomous AI agents have demonstrated remarkable capabilities in code generation, planning, and optimization. However, in chip design, raw agentic intelligence is not enough. The performance, reliability, and cost efficiency of agent-driven workflows depend critically on how deeply they internalize domain knowledge, from high-level architectural intent and design-flow structure to low-level algorithmic heuristics and hardware constraints. This talk will discuss our findings of how domain knowledge transforms generic agents into high-performance chip-design systems. I will discuss domain-aware bootstrapping, expert-inspired subagent orchestration, and self-evolving optimization loops across three core areas in chip design: EDA tool R&D (software), RTL designs (hardware), and technologies (VLSI).
Bio
Prof. Cunxi Yu is an Assistant Professor at the University of Maryland, College Park. His research focuses on novel algorithms, systems, and hardware designs for computing and artificial intelligence. His work has been recognized with a Best Paper Award at DAC (2023), a Best Paper Award at ASPLOS (2025), Best Paper Nominations at DAC, ICCAD, and ASP-DAC, as well as the NVIDIA Academic Research Award and the NSF CAREER Award. He also holds a Visiting Professor position at NVIDIA Research. Prof. Yu earned his Ph.D. from the University of Massachusetts Amherst in 2017.