PhD students win major awards at Design Automation Conference

Members of the eLab Research Group in the Department of Electrical and Computer Engineering at the University of Wisconsin-Madison received major awards at DAC 2026, The Chips to System Conference, considered the oldest and most prestigious conference in electronic design automation and semiconductor-system design. The conference was held in Long Beach, California, in late July 2026.

PhD students Alish Kanani, advised by Gene Amdahl Professor Umit Ogras, and Lukas Pfromm, co-advised by Ogras and Dugald C. Jackson Assistant Professor Eric Tervo, were co-first authors on a paper that received the ACM Transactions on Design Automation of Electronic Systems (TODAES) Best Paper Award. TODAES is one of the leading journals in the field, and only a single paper published during the year receives the honor. The selection considers originality, timeliness, potential impact, and overall quality. The paper is titled “MFIT: Multi-Fidelity Thermal Modeling for 2.5D and 3D Multi-Chiplet Architectures.”

One of the limiting factors in scaling modern and emerging computing systems is thermal management. When high-power chiplets are positioned next to one another or stacked vertically, their heat can cause performance and reliability problems.

That’s why system architects use high-fidelity thermal simulations to evaluate their chip designs. However, current simulations can take days to run, making them impractical for evaluating the thousands of possible architectures or supporting runtime decisions.

MFIT gives system designers fast and accurate ways to predict and manage heat in 2.5D and 3D chiplet systems. Its high-fidelity finite-element models support detailed chip and package development; faster thermal RC models enable architecture exploration, cooling design and design-space optimization; and millisecond-scale discrete state-space models provide real-time thermal information for workload scheduling, dynamic power management and hotspot prevention.

The open-source project allows researchers and engineers to modify, validate and integrate MFIT into their own design workflows and is already being used in multiple academic labs and industry research groups.

Other UW-Madison authors include Tervo, Ogras and MS student Parth Solanki.

Alish Kanani also received another significant honor at the conference, taking first place at the DAC PhD Forum. Kanani won for his dissertation research, titled “Cross-Layer Design of Heterogeneous Multi-Chiplet Systems for Efficient and Scalable AI Inference.”

Kanani’s PhD research develops a unified methodology that connects three stages of heterogeneous chiplet-system design: modeling physical and thermal limits, specializing hardware for different AI workload phases, and dynamically scheduling workloads across heterogeneous chiplets. The work, developed with his PhD advisor Ogras, has practical applications for developing faster, more energy-efficient, and thermally safe AI accelerators for large language models and other neural-network workloads.

The DAC PhD Forum is a highly competitive international competition, hosted by the Association for Computing Machinery (ACM) Special Interest Group on Design Automation (SIGDA) and the IEEE Council on Electronic Design Automation (CEDA), with just 28 PhD candidates selected to present their dissertation research in 2026.

Featured image: From left to right, UW-Madison DAC 2026 attendees Miao Sun, Zhengxiong Li, Jiahao Lin, Alish Kanani, and Umit Ogras. Credit: Submitted