Graduate student’s thesis work nets her spots at top conferences

V Pulavarthi

Vaishnavi Pulavarthi’s master’s thesis on using large-language models in hardware design and verification netted her a paper at a top-tier computational linguistics conference and a special session at another premier electronic design automation conference.

Pulavarthi successfully defended her thesis in June 2024. She worked in IT as a systems operations analyst and programmer at Wells Fargo for three years before joining Assistant Professor Debjit Pal’s research group in 2023. She recognized her passion for electrical digital hardware design, motivating her to choose UIC for her graduate studies.

Pulavarthi implemented machine learning into system design, using a handful of large language models (LLMs), such as ChatGPT, LLaMa3, and Code-LLaMa. She wanted to see how these models would understand or approach hardware design when given prompts.

“I created a data set and used LLMs to find out if they understand what an assertion is,” Pulavarthi said. “An assertion is a mathematical expression in Boolean algebra capturing temporal behavior of a hardware design. Is it giving me a random answer, or does it understand what’s going on in that mathematical expression?”

Pulavarthi implemented a technique to measure what the LLMs understood and then trained them to generate mathematical expressions rather than provide random results, known as hallucinations. She found that using AI in hardware verification holds promise but still needs more fine-tuning.

Pulavarthi’s paper was accepted at the 2025 annual conference of the North American Chapter of the Association for Computational Linguistics (NAACL), a top conference in languages and linguistics.

Her work also set the theme of a special session on generative AI for assertion generation in the 2025 Design Automation and Test in Europe (DATE), another top conference in electronic design automation.

Pulavarthi is actively seeking her next role in the industry.