Join iMAGiNE and the Chandra Family Department of Electrical and Computer Engineering for the iMAGiNE Symposium 2026. We are excited to be joined by AMD CTO, Mark Papermaster for our Keynote and Fireside Chat.
The iMAGiNE Consortium provides tools, methodologies, and knowledge for engineering the machines that support intelligent applications, from the smallest circuits to the largest systems. Prospective industry members and The University of Texas at Austin faculty, students, and alumni are welcome to attend!
Registration & Networking Breakfast
8:30 - 9:00 AM, EER 0.804 (Mulva Foyer)
Welcome Remarks
9:00 - 9:15 AM, EER 0.904 (Mulva Auditorium)
Roger Bonnecaze
Dean, Cockrell School of Engineering
Diana Marculescu
Chair, Chandra Family Department of Electrical and Computer Engineering
Founding Director, iMAGiNE
Opening Keynote
9:15 - 10:00 AM, EER 0.904 (Mulva Auditorium)
Mark Papermaster
CTO, AMD
Mark Papermaster, CTO and Executive Vice President at Advanced Micro Devices (AMD), has been instrumental in AMD’s revival over the past 15 years. He spearheaded the overhaul of AMD’s engineering processes and the creation of the award-winning “Zen” high-performance x86 CPU family, high-performance GPUs, and AMD’s modular design approach, Infinity Architecture.
A holder of a B.S. from The University of Texas at Austin and an M.S. from the University of Vermont, both in electrical engineering, Mark serves on several boards, including the Global Semiconductor Alliance Board of Directors, IEEE Industry Advisory Board, Cockrell School of Engineering Advisory Board, the UT Austin President’s Austin Innovation Board, and Purdue University Semiconductor Degrees Leadership Board.
-- Enabling Efficient AI: Together We iMAGiNE --
AI is rapidly becoming pervasive—from the data center to the edge, to scientific research labs—unlocking new applications, productivity gains, and national-scale infrastructure. But this growth is colliding with a power wall. In this keynote, AMD CTO Mark Papermaster will share how constant innovation and holistic design, spanning chiplet-based modularity, rack-scale system design, and co-optimized software, can deliver step-function gains in compute efficiency and scalability. He will also look ahead to emerging frontiers, including new forms of acceleration like quantum, and highlight AMD’s role in advancing open ecosystems, embedded AI platforms, and deep partnerships with academia and industry.
Fireside Chat
10:00 - 10:30 AM, EER 0.904 (Mulva Auditorium)
Diana Marculescu, Mark Papermaster
Panel
10:30 - 11:30 AM, EER 0.904 (Mulva Auditorium)
-- Efficient AI at Scale --
As AI systems proliferate across the edge-cloud continuum, achieving efficiency at scale has become a defining challenge. This panel explores how AI performance and accuracy, energy consumption, and total cost of ownership intersect across edge devices and cloud data centers, where scaling AI often demands enormous infrastructure and power. The panel also uncovers a core tension of efficiency: as AI systems become more efficient and accessible, Jevons Paradox suggests that usage may expand dramatically, amplifying demands on datacenter and edge infrastructure while raising deep implications for privacy and trust. Panelists will debate how to balance efficiency-driven scale with safeguards that ensure trustworthy AI deployment across the edge and cloud.
Moderator
Mattan Erez
Professor, Chandra Family Department of Electrical and Computer Engineering
Panelists
Eric Van Hensbergen
Fellow, ATG Leadership Team, Research, Arm
Jean Anne Incorvia
Associate Professor, Chandra Family Department of Electrical and Computer Engineering
Lizy John
Professor, Chandra Family Department of Electrical and Computer Engineering
Denis Foo Kune
Senior Security Researcher, Artificial General Intelligence and Responsible AI, Amazon
Radu Marculescu
Professor, Chandra Family Department of Electrical and Computer Engineering
Michael Schulte
Senior Fellow Design Engineer, AMD
Student Demos
11:30 AM - 12:00 PM, EER 0.904 (Mulva Auditorium)
Join us for this session showcasing hands-on innovations in efficient AI systems for real-world deployment! The demos feature topics ranging from energy-harvesting platforms running AI for personal health and wellness, to small-footprint, low-power AI systems for embodied intelligence and robotics operating under tight energy and resource constraints.
Moderator
Edison Thomaz
Associate Professor, Chandra Family Department of Electrical and Computer Engineering
Demos
AI Hardware Platform for Edge Computing
Working with Sandia and ASU on a CMOS+X platform for AI acceleration using nonvolatile memory crossbar arrays. We can demo it working with RRAM and maybe with MRAM if it is ready in time.We did a short movie showing how we can write the desired conductance states into the RRAM array through the 2.5D interposer. We can do some demos like this.
Batteryless Wearable AI: Toward Scalable, Invisible Sensing
We will demonstrate a working prototype of 5-7 batteryless wearable sensors that perform real-time human activity recognition using only harvested ambient light. The accelerometer-based nodes operate entirely without batteries, running energy-aware ML inference under intermittent indoor lighting.
While the current system is a proof of concept, the broader vision aligns directly with Efficient AI at scale. Batteries are a key barrier to large-scale deployment. They limit lifetime, require maintenance, and prevent dense embedding of sensors into everyday environments. By eliminating the battery, we enable self-powered, maintenance-free sensing that can scale to hundreds or thousands of nodes embedded into clothing, furniture, and built spaces.
This work reframes scaling AI not just as scaling models, but scaling sustainable deployment. The demo will show live activity recognition powered solely by room light and illustrate how sensing can become part of the fabric of our environment.
Detection of Ingestive Behavior System (DIBS)
The DIBS device is a small hardware platform designed and optimized for eating detection in-the-wild via inertial sensing on the jaw. It consists of a PCB containing a BLE equipped microcontroller, 6-axis IMU, and LiPo battery. The width and height of this form factor is roughly the size of the diameter of a penny allowing for it to be adhered to the outside of the jaw bone comfortably. The system is designed to connect to a companion smartphone app where the data can be offloaded for processing depending on the application. The battery life of the device is roughly 20-24 hours during active BLE transmission. We applied machine learning techniques to the 6-axis inertial data for binary eating detection and have achieved performance over 91% F1 score. We will have the device to demonstrate as well as some figures describing the device power consumption and machine learning model performance as the model size decreases.
Networking Lunch & Student Poster Session
12:00 - 1:30 PM, EER 0.804 (Mulva Foyer)
Lunch provided through the generous support of Amazon Science Hub.
Members Only Session
1:30 - 3:00 PM, EER 0.806/0.808 (Multipurpose Room)