CAREER: Human-AI Collaboration to Debug Analog Circuits

Today, analog circuit debugging is a daunting task because of its manual nature and heavy reliance on experience. A manual root cause search is time-consuming, which creates a profitability issue for the semiconductor industry by delaying time-to-market and time-to-revenue. Over-reliance on human subject matter experts also poses a workforce challenge: if experienced experts leave or retire before new engineers are fully trained, critical knowledge may be lost. This research seeks to address both problems by building an artificial intelligence (AI)-based virtual expert to help people think. This project also investigates how to provide privacy guarantees to debugging queries when using cloud-based, closed-source large language models. This research has the potential to increase the economic competitiveness of U.S. semiconductor companies by boosting productivity through automation, leading to cheaper and more advanced sensors, smartphones, and mobile devices for everyone. Privacy research will also protect company intellectual property, another competitive advantage of the U.S. semiconductor industry. This project aims to automate analog circuit debugging through human-AI collaboration. Specifically, this research focuses on the architectural and algorithmic foundations within various AI technologies for analog experience accumulation, representation, and delivery to novices during human-AI collaborative debugging to expedite bug localization. The project tasks include designing retrieval-augmented generation for single-step root cause recommendation, multi-agent systems with conflicting goal management (hypothesis generation vs. problem-space reduction) to balance intuition with pragmatism during multi-step debugging, and differential privacy-based security guarantees for bug-related information. This CAREER project tightly integrates research and education by using research artifacts (e.g., AI copilots) to improve debugging education at both the college and K-12 levels. Educational activities will also use and generate real-life data for research. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2543901 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT,01003031DB NSF RESEARCH & RELATED ACTIVIT,01002930DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: John Hu | Institution: Oklahoma State University, STILLWATER, OK | Award Amount: $357,497 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2543901 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2543901.html