Mark of Agentrys AI

The semiconductor industry has achieved extraordinary productivity gains over the past half century through successive waves of design automation, from circuit simulation and logic synthesis to timing, power, and physical implementation engines. Yet today’s chip design process still depends heavily on human engineers to translate specifications, write and debug RTL and verification code, analyze reports and logs, tune design recipes, close coverage, and optimize PPA across complex tool flows.

This talk introduces Agentic Design Automation (ADA) as the next major step in chip design productivity. Unlike traditional EDA tools that automate isolated tasks, ADA combines frontier models, custom domain models, agent-native tools, and existing EDA engines into closed-loop agents that can reason, code, debug, analyze, and optimize across the chip design workflow. The talk will discuss how agents become substantially more capable when equipped with the right tools: causal-graph debugging for verification failures, e-graph rewriting for RTL optimization, log-structure extraction, solver improvement, RTL-specialized models, and self-improving multi-agent workflows.

Through recent examples in verification debug, RTL optimization, SAT-solver evolution, RTL code generation, and autonomous agent improvement, the talk will show how agentic workflows are beginning to move beyond one-off LLM assistance toward reusable, measurable, and continuously improving design automation systems. Results such as significant RTL optimization gains, improved solver performance, high pass rates on RTL coding/debug benchmarks, and autonomous agent evolution illustrate the potential of this direction.

The central message is that the future of chip design will not be defined only by better models, but by better agent systems: agents connected to specialized models, design databases, EDA tools, verification engines, and organization-specific knowledge. As ADA matures, the role of designers will shift from manually executing every design task to developing, supervising, and improving the agents that design chips. In this future, agents design chips, and designers develop the agents.

Biography:

Mark Ren is the Founder and CEO of Agentrys, building Agentic Design Automation to redefine how chips are designed and built. He is an IEEE Fellow, and a recognized leader in EDA and AI for chip design, with 26 years of R&D experience across IBM Research and NVIDIA Research. At IBM, he advanced core digital implementation technologies including placement, ECO synthesis, and datapath optimization, work that contributed to the IBM Corporate Award for improving microprocessor design productivity. At NVIDIA, he built and led a 10+ person design automation research team, helped establish industry leadership in GPU-accelerated EDA, drove large-scale adoption of AI in EDA, and led ChipNeMo, the first industrial LLM initiative for chip design.

Mark’s work spans digital, analog, and PCB design automation. He has received 2 DAC Best Paper Awards and more than 10 additional best paper awards or nominations across major EDA conferences. He has delivered keynotes for Cadence, Synopsys, TSMC, Intel, and Silicon Labs; taught AI-for-chip-design tutorials at DAC and Hot Chips; chaired the LLM Aided Design conference; and served on ICCAD and ISPD committees. His work has been featured in WIRED, Fortune, and EE Times. His expertise includes GPU‑accelerated and AI‑driven EDA, LLMs and agent systems for chip design, physical design and logic synthesis, PPA optimization, and translating research into production systems.