Events Listing

Professor Samuel Palermo at ECE spring 2026 colloquium

Growing datacenter bandwidths datacenters requires optical transceivers operating at high data rates. Further increases in bandwidth density is possible with Wavelength-division multiplexing, which architectures based on silicon photonic microring modulators (MRMs) inherently enable. This talk covers high-speed PAM4 transmitter and receiver front-ends implemented in a 28nm CMOS process that are co-designed with these silicon photonic optical devices. The transmitter utilizes an optical DAC approach with two PAM2 AC-coupled pulsed-cascode high-swing output stages to drive the MRM MSB/LSB segments with a 3.42V ppd at 80Gb/s PAM4. The receiver consists of a transimpedance amplifier with sub-Nyquist bandwidth for low input-referred noise and a subsequent continuous-time linear equalizer for bandwidth recovery. Efficient clocking is realized with an LC-oscillator-based quarter-rate digital clock and data recovery system. The RX achieves 100Gb/s PAM4 operation with −6.4 dBm sensitivity.

Professor Andrew Kent at ECE spring 2026 colloquium

Magnetic tunnel junctions (MTJs) are widely used as nonvolatile memory elements, but they can also serve as controllable, high-rate sources of random bits. In this talk, I will describe experimental studies of perpendicularly magnetized MTJs that are magnetically stable at room temperature. Instead of relying on spontaneous thermal magnetization fluctuations (superparamagnetism), stochastic behavior is generated on demand by actuating the device with nanosecond electrical pulses in the ballistic spin-transfer regime. This approach enables precise control of the switching probability. I will present measurements showing high-rate (up to 100 MHz/MTJ), reproducible generation of random bit streams and random telegraph noise. By interfacing individual pMTJs with custom electronics and a field-programmable gate array (FPGA), we generate truly random numbers that pass the full NIST statistical test suite with no post-processing. I will also show how such actuated stochastic MTJs (A-sMTJs) can be electrically connected in simple circuits to generate tunable, circuit-mediated interactions that map onto effective Ising couplings. Finally, I will discuss the potential of stochastic MTJs or physics-inspired computing systems, including their use for solving combinatorial optimization problems.

Professor Rosana Esteller at ECE 2026 Spring Colloquium

Neuromodulation is shifting from conventional open-loop therapy toward intelligent systems that integrate sensing, signal processing, modeling, and adaptive stimulation. This talk frames neuromodulation as an engineering problem in which biomarker algorithms estimate physiologic state, closed-loop systems use feedback to adjust therapy, and dynamic stimulation strategies expand control beyond conventional static pulse trains. Drawing on translational work in implantable neurotechnology, the presentation will discuss the design of biomarker-driven algorithms, practical constraints in real-time closed-loop implementation, and the biophysical rationale for time-varying stimulation. Together, these approaches highlight how engineering can enable more precise, personalized, and effective neuromodulation therapies.

Professor Zhichao Cao at ECE 2026 Spring Colloquium

In this talk, Professor Zhichao Cao will explore the symbiotic relationship between AI and data systems, showing how Large Language Models (LLMs) both automate data systems and drive new data system designs. He will present two representative projects: (1) StorageXTuner, the first LLM agent–driven framework that automatically tunes performance for diverse data systems such as RocksDB, LevelDB, MySQL, and CacheLib, outperforming traditional tuning methods; (2) M2Cache, a system that makes LLM inference sustainable and accessible on outdated or low-end hardware through a co-design of dynamic mixed-precision inference and a predictive multi-level cache across HBM, DRAM, and SSDs, greatly improving efficiency and reducing carbon footprint. Finally, Cao will conclude with his vision for future research at the intersection of AI and data systems.

Registration closes at 12 CST on Monday, March 9, 2026

Join CBI's 2025-2026 Tomash Fellow Sam Franz Doctoral Candidate, History and Sociology of Science, University of Pennsylvania, presenting his paper Calculating Knowledge: Computing, Capitalism, and the Modern University, 1945–1990.

This virtual event is free and open to the public.

Title: Calculating Knowledge: Computing, Capitalism, and the Modern University, 1945–1990

Abstract: This talk presents my dissertation, Calculating Knowledge, which argues that between 1945 and 1990 U.S. universities made industrial the production of scientific knowledge by embedding computing as both a managerial technology and an intellectual practice. Focusing on the central institution of the computing center, I show how these technologies and their institutional formations reorganized research, teaching, and administration around the practical and intellectual problems that confronted their operation. In the process, universities became key sites where the automation of intellectual labor in the form of computing machinery and emerging ideas about the “knowledge economy” converged. Drawing on archival research across public and private institutions, I show how computing centers and eventually computer science departments reshaped academic labor and redefined the university as an engine of economic growth. By placing the university itself at the center of the history of computing, the project reframes postwar transformations in work, capitalism, and higher education and offers a longer history for contemporary debates about AI and intellectual labor.

Alejandro Ribeiro at ECE 2026 Spring Colloquium

Layers of graph neural networks (GNNs) involve linear operators that admit scalar representations on the graph's spectrum. We use these spectral representations to investigate the response of GNNs to graph perturbations (stability) and scaling towards continuous limits in the form of manifolds or Graphons (transferability). We uncover that stability and transferability are determined by the Lipschitz constant of these frequency responses. Since Lipschitz constants are related to spectral discriminability, these analyses uncover fundamental stability and transferability vs discriminability tradeoffs. In particular, GNNs that attempt to discern features aligned with graph eigenvectors associated with larger eigenvalues are more unstable and more difficult to transfer at scale. Scalar (or low dimensional) spectral representations are common to any architecture that involves the composition of an operator. This includes GNNs as well as standard convolutional neural networks (CNNs) in time and space along with other less common information processing architectures. We use tools of algebraic signal processing to show that all these architectures share analogous stability and transferability properties. Materials available at the Graph Neural Networks website.

Matteo Cercola at ECE 2026 Spring Colloquium

Professor Mahdi Bayat at ECE spring 2026 colloquium

Artificial intelligence and deep learning in particular is rapidly changing the field of medical imaging. In this talk I will give an overview of some of the important applications of deep learning in medical imaging with some examples from the real world clinical and biomedical applications. I will also discuss AI-ready imaging informatics as the backbone for preparation of training data as well as providing the infrastructure for clinical deployment. 

Professor Ying Wang at ECE 2026 Spring Colloquium

Symmetry breaking in quantum and two-dimensional materials provides a unifying framework for emergent electronic properties and new device functionalities. When structural or electronic symmetries are broken, quantum materials can host strong correlations, nontrivial quantum geometry, and spontaneous electric dipoles, leading to nonlinear responses beyond conventional semiconductors.In the first part of this talk, I will show how symmetry breaking in correlated Weyl semimetals gives rise to exceptionally strong nonlinear transport, and how this large nonlinearity enables a highly sensitive terahertz detection scheme.In the second part, I will discuss symmetry breaking in layered two-dimensional materials, where interlayer sliding induces switchable electric dipoles. This sliding ferroelectricity enables ferroelectric tunnel junctions with giant resistance contrast, ultralow switching energy, and high endurance, offering a new platform for energy-efficient nonvolatile memory devices.

Professor Yogatheesan Varatharajah at ECE 2026 spring colloquium

Recent advances in wearables, brain implants, and sensing technology have enabled us to design systems that continuously monitor patients' brain health and ascertain individualized treatments for neurological diseases. However, there is a lack of efficient methods that translate continuous physiological data streams into meaningful biological models of underlying diseases, relate them to existing clinical knowledge and biomarkers, and provide actionable treatment parameters. Machine learning (ML) holds great promise in tackling these challenges; however, the mainstream black-box-ML approaches have proven to be untrustworthy because of label inconsistencies, spurious correlations, and the lack of deployment robustness. My goal is to ensure trustworthiness in ML for healthcare, particularly neurology, via a novel framework known as “Domain-guided Machine Learning” or “DGML” that merges machine learning with clinical domain expertise. In this talk, I will discuss the need for trustworthy ML in healthcare, how to leverage clinical domain knowledge to engineer trustworthy ML models, and several real-world applications of DGML in neurological care and decision making.