You’re awakened by the alarm on your smartphone. Deciding to have a quick bite before starting your day, you open the freezer and grab a microwave meal to heat up. While waiting for the microwave to finish, you check your text messages and ask ChatGPT a quick question about how to dress for today’s weather.
It’s an ordinary morning, but you may not realize that the devices and systems powering your daily life depend on semiconductor integrated circuits, commonly referred to as microchips or chips. There are several types of chips, including digital, analog, and mixed-signal.
Digital chips benefit from highly automated design tools, whereas analog chip design still relies heavily on manual processes. This reliance results in slower, more expensive, and more challenging scalability for analog chip design.
Researchers have been working on analog design automation for decades, but progress has been limited, partly due to the lack of shared datasets, standardized benchmarks, and scalable research infrastructure. This issue has become increasingly urgent with the rapid growth of artificial intelligence (AI) and machine learning (ML), which rely on large-scale data and reproducible evaluation environments.
Electrical and Computer Engineering Professors Weidong Cao and Tian Lan have been awarded a grant from the National Science Foundation to support their work addressing this need and build on Cao’s existing research in AI-driven chip design. Through this grant, they will create AnalogEDA-Hub, an open platform for AI-driven research in analog electronic design automation (EDA) at scale.
“Analog EDA research is currently fragmented,” Cao explained. Researchers often develop their own datasets, circuit environments, evaluation methods, and design workflows. This makes it difficult to compare approaches fairly, reproduce results, or transfer a method from one circuit or design task to another,” he said.
“AnalogEDA-Hub is intended to provide a common foundation for the community. It will bring together multimodal datasets, representative analog circuits, standardized design environments, benchmarking capabilities, and learning-based tools in one open infrastructure,” said Cao.
Cao and Lan plan to directly engage GW students in their work, both in their lab and through external outreach. Graduate and undergraduate students will collaborate with Cao and Lan to develop datasets, AI and ML methods, design automation algorithms, benchmarking environments, and analog circuit applications.
They also plan to utilize AnalogEDA-Hub as an educational platform. This infrastructure will provide standardized, accessible design environments, enabling students to learn about analog chip design, EDA, and AI through hands-on projects rather than relying solely on traditional classroom instruction. By exposing students to semiconductor design and AI-driven engineering, the initiative aims to create a broader pathway into the semiconductor workforce.
“Ultimately, our vision is to make analog chip design significantly more scalable, accessible, and intelligent, helping move the field from a largely handcrafted process toward a new generation of AI-enabled design methodologies,” Cao concluded.