- Python
- Introduces the first agentic approach to automate the design rule checking engineering change order (DRC ECO) process.
- Maps DRC repair into a skill-file evolution task, enabling agents to iteratively learn and refine layer-specific repair strategies.
- Python
- Open-sourced a benchmark for evaluating LLMs on chip layout DRC detection and repair tasks.
- Provides a standardized, containerized pipeline to assess how well models identify and fix design rule violations.
- Python
- Open-sourced the diffusion-based physical design framework that generates synthetic circuit heatmaps for ML-based EDA research.
- Open-sourced a collection of over 23,000 synthetic circuits and heatmaps for machine learning-based EDA research.
- Python
- Open-sourced the framework that integrates the script-generating LLM with the physical design tool.
- Combined prompt engineering with the physical design tool’s feedback to iteratively generate the tool script.
- Verilog/Tcl/Python/C++
- Used C++ and SWIG to create Python APIs in OpenROAD, enabling gate sizing operations using OpenROAD’s Python APIs.
- Developed Python scripts for examples and evaluations for the contest.
- Used EDA tools to synthesize netlists with the ASAP7 library and perform placement and routing.
- Python
- Open-sourced the LLM for generating scripts for the physical design tools and answering questions related to the physical design tools.
- Python
- Open-sourced the first dataset of physical design tool scripts for LLM-based physical design research.
- C++/Python
- Developed STA-related Python API for OpenROAD to provide flexibility in ML-EDA.
- Created demos on using the OpenROAD Python Interface and using Nvidia’s CircuitOps to build data pipelines for ML-based EDA research.
- Tcl/Python
- Developed an ML-friendly data infrastructure to generate datasets for ML-EDA applications.
