The Ph.D. student was one of seven students globally to receive the newly established IEEE LLM-Aided Design Fellowship, awarded for research using large language models to automate hardware design.
(text and background only visible when logged in)
(text and background only visible when logged in)
Georgia Tech School of Electrical and Computer Engineering (ECE) Ph.D. student Stefan Abi-Karam was named one of seven recipients of the inaugural IEEE Large Language Model Aided Design (LAD) Fellowships.
The award supports graduate students exploring how artificial intelligence can transform the way computer chips are designed.
Abi-Karam received the fellowship for his proposal, “Enabling Rapid Optimized Domain Acceleration With Agentic High-Level Synthesis.”
The research focuses on using large language models (LLMs)—the same technology behind tools like ChatGPT—to design specialized chips called hardware accelerators. These accelerators make demanding computational tasks faster and more efficient, from autonomous driving and robotics to medical imaging and efficient AI processing.
“In short, we want to go from a natural language prompt and application specification to an efficient design running on real hardware using an end-to-end automated LLM hardware design agent,” Abi-Karam said. “The approach is a promising roadmap to reach that goal.”
Today, chip design often relies on low-level coding languages that are difficult for LLMs to handle.
Abi-Karam’s approach uses high-level synthesis (HLS), which starts with more familiar languages like C++ and converts them into the lower-level code needed for manufacturing. This higher-level starting point makes it easier for AI to optimize designs and improve performance without getting bogged down in technical details.
Through the fellowship, he plans to further explore HLS and integrate the many different HLS tools, both commercial and the custom academic tools from Assistant Professor Callie Hao’s Software/Hardware Co-Design for Intelligence and Efficiency Lab (SHARC), into these LLM HLS design agents.
By the end of the fellowship, Abi-Karam hopes to create an automated agent that can produce a fully functional design, ready to run on real hardware.
He is in the fourth year of his Ph.D. studies at Georgia Tech, with Hao as his advisor. On top of his studies, Abi-Karam is also a full-time research faculty member through Georgia Tech Research Institute (GTRI) in the Cybersecurity, Information, Communication, Command and Control and Software Systems (CIPHER) lab division.
Related Content
Sharc Lab’s 3D Chip Design Research Wins Best Paper Award
Ph.D. students Ismael Youssef and Hang Yang introduced LaZagna, a tool for exploring billions of 3D field-programmable gate array design options, earning top honors at the International Conference on Computer-Aided Design.
Smart Sensor System Targets Global Gaps in Menstrual Health Care
The initiative, co-led by Ali Adibi and Ashkan Zandi, is developing sensor technology in menstrual products to detect heavy bleeding, addressing a condition often overlooked until it causes severe health consequences.