Can a Language Model Be Your Hardware Security Co-Designer and Verifier?

Play Video

Speaker 

Dr. Hadi Kamali is an assistant professor in the Department of Electrical and Computer Engineering (ECE) at the University of Central Florida.

Abstract

Recent advances in large language models (LLMs) have unlocked their potential not only to assist in hardware design but also to be actively used for its verification and assurance. This talk will demonstrate how coupling learned language representations with domain-specific structural knowledge, such as syntax trees, control-flow graphs, and dataflow graphs, can lead to reliable hardware-generation LLMs that are both functionally correct and structurally sound. We also show how decoding strategies in LLMs can further reduce design flaws and ensure compliance with security and verification requirements. This convergence of generation, structural awareness, and automated verification points toward a new paradigm: LLMs as collaborative agents in hardware design pipelines, capable of co-designing secure architectures and serving as early-stage verifiers. In this talk, we explore how such models move us toward trustworthy, AI-augmented hardware creation, where security is embedded from the first line of code.

Biography

Dr. Hadi Kamali is an assistant professor in the Department of Electrical and Computer Engineering (ECE) at the University of Central Florida. He received his Ph.D. degree from the Department of ECE at George Mason University in 2021. His research delves into hardware security with a particular focus on exploiting IP protection techniques, design-for-trust for VLSI circuits, and CAD frameworks for security (design-for-security). His research contributions include the authorship of two books, two book chapters, five issued/pending patents, and 60+ publications featured in top-tier journals and conferences. Dr. Kamali’s research has received recognition through several awards, including nominations and the recipients of the Best Paper Award in ICCAD’19, ISVLSI’20, ICCAD’20, DCAS’20, HOST’22, DATE’23, ISEC’25, and ASP-DAC’25.