Tuesday 26 August 2025
A team of researchers has developed a framework that enables electronic design automation (EDA) workflows to be automated using large language models (LLMs). The system, called AutoEDA, aims to simplify and speed up the complex process of designing integrated circuits.
The EDA workflow is a crucial step in creating modern electronics, from smartphones to computers. It involves multiple stages, including synthesizing code, placing components on a chip, and optimizing performance. Traditionally, engineers write custom scripts in languages like TCL to control these stages, but this approach can be labor-intensive and prone to errors.
AutoEDA uses LLMs, which are trained on vast amounts of text data, to understand the design intent and generate the necessary code for each stage. The framework is designed to work with popular EDA tools, such as Synopsys Design Compiler and Cadence Innovus.
The researchers used a benchmark of 100 diverse prompts and corresponding TCL mappings to evaluate AutoEDA’s performance. The results showed that the system achieved a 2.4-fold improvement in code quality compared to in-context learning baselines, while reducing output token usage by over 75%.
AutoEDA is not only faster but also more accurate than traditional scripting methods. This is because LLMs can learn from vast amounts of data and adapt to new situations, allowing them to generate code that is tailored to specific design requirements.
The framework also addresses the limitations of prior approaches, which often relied on direct script generation or fine-tuned models with poor generalizability. AutoEDA’s modular microservice backend enables it to work seamlessly with different EDA tools and workflows, making it a versatile solution for various applications.
The development of AutoEDA has significant implications for the electronics industry. By automating the EDA workflow, designers can focus on higher-level design decisions rather than tedious scripting tasks. This could lead to faster time-to-market, reduced costs, and improved product quality.
As the demand for complex electronic systems continues to grow, the need for efficient EDA workflows will only increase. AutoEDA’s innovative approach using LLMs has the potential to revolutionize the way we design and develop modern electronics.
Cite this article: “Automating Electronic Design Automation with Large Language Models”, The Science Archive, 2025.
Electronic Design Automation, Large Language Models, Integrated Circuits, Computer-Aided Design, Code Generation, Synopsys, Cadence, Tcl, Microservices, Natural Language Processing.
