The recipients of the 2020-2021 Graduate School Doctoral Dissertation Fellowship are Alireza Sadeghi (advisor: Prof. Georgios Giannakis), Karthik Srinivasan (advisor: Prof. Beth Stadler), Masoud Zabihi (advisor: Prof. Sachin Sapatnekar), and Kaveh Khilji (advisor: Prof. Tony Low)
Alireza Sadeghi is a recipient of the fellowship for his doctoral research titled, “Scalable Learning Robust to Uncertainties with Applications in Cyber-Physical Systems.” He is working under the guidance of Prof. Georgios Giannakis.
Alireza’s research interests span across several areas: artificial intelligence, machine learning, signal processing, and optimization with applications in networks including smart power networks, wired and wireless networks. In his doctoral research, Alireza addresses some of the challenges facing machine learning. Currently, machine learning algorithms are vulnerable to adversarially manipulated input data, and to uncertain environments. This discourages their use in safety-critical applications. Besides, such algorithms often rely on the premise that training and testing data are drawn from the same distribution, which may not hold in practice.
Alireza’s work targets these challenges, and builds learning models that are robust to uncertainties arising from, for instance, distributional mismatches. By wedding innovative machine learning tools, with recent advances in stochastic function interval estimation, robust optimization, control, networking, and communications, he develops scalable and robust algorithms with applications in cyber-physical systems.
Alireza Sadeghi earned his bachelor’s degree from Iran University of Science and Technology, Tehran, in 2012, and his master’s degree from University of Tehran in 2015 (both in electrical engineering). He is currently pursuing his doctoral degree with the Department of Electrical and Computer Engineering.
In 2015, he was a visiting scholar with the Department of Information Engineering (DEI) at the University of Padua, Padua, Italy. Previously, he was a recipient of the ADC Fellowship awarded by the Digital Technology Center at the University of Minnesota Twin Cities, and the Student Travel Awards from the IEEE Communications Society and the National Science Foundation.
Karthik Srinivasan is a recipient of the Doctoral Dissertation Fellowship awarded by the University’s Graduate School for his research titled, “Magneto-Optical Isolators – The “Missing-Link” in Integrated Photonics.” He is conducting his research under the guidance of Prof. Bethanie Stadler.
Karthik’s work is primed for the future of the computational world, as it moves away from pure electronics towards using photons, spins, and magnons for solving emerging computational problems. His primary research interest is in process development and characterization of novel magneto-optic materials with unique gyrotropic and magnonic properties that can be used for the design of photonic integrated circuits and high frequency microwave filters.
While a fully integrated photonic circuit can perform computations significantly faster than a current day electronic chip, the challenge remains that such a photonic circuit is impeded by the lack of chip-scale optical isolators. These isolators allow for the unidirectional propagation of light which is critical to the protection of on-chip lasers from destabilizing reflections. Karthik is working on the development of exotic magneto-optical materials for Si-integrated isolators that can manipulate light regardless of an external magnetic field. He is currently focused on ways to increase the gyrotropy of cerium doped terbium iron garnet (Ce:TbIG), and to investigate material properties that support magnetless isolation.
One of the key outcomes of Karthik’s research so far is that waveguide isolators fabricated with this new garnet match the mode and dimensions for on-chip lasers. In terms of specific outcomes, these isolators allow for magnetless isolation and increase up to 40 times in device density, which translates to at least 85000 devices per square inch on a photonic integrated circuit.
What’s next for Karthik? He intends to continue working on downsizing waveguide isolators. And the next step to that is the development of garnets with positive Faraday rotation to complement existing negative Faraday rotating garnets. Successful development of such garnets would mean a 50% reduction in device dimensions.
Currently, the absence of an on-chip laser-isolator pair has been a bottleneck even as the photonics community is making significant strides in the development of components such as modulators, circulators, and logic-gates. However, the development of a “ready-to-integrate” optical isolator that is foundry friendly and favorable for industry adoption could change that. Karthik’s research brings us closer to the goal, while simultaneously contributing knowledge to the field.
Karthik Srinivasan earned his bachelor’s degree in electronics and communication engineering from Anna University, in the southern Indian city of Chennai. He moved to Minneapolis in 2016 and earned his masters degree in electrical engineering from the University of Minnesota Twin Cities in Spring 2019. His research lies at the intersection of photonics, magnetism and materials, and addresses the challenges of data storage and computation needs for high-speed high-volume processes.
Besides the highly competitive doctoral dissertation fellowship, Karthik is also the recipient of a travel Award by the IEEE Magnetics Society (2019), and a fully sponsored IEEE magnetics summer school in Quito, Ecuador (2018; he was one of 70 students selected from around the world). He is the Vice-Chairperson for the IEEE Magnetics Society chapter for Twin Cities, MN; he has held the position for three consecutive years now.
Masoud Zabihi’s fellowship winning doctoral research is titled “In-Memory Processing Using Spintronics Computational RAM (CRAM).” He is working on his dissertation under the guidance of Prof. Sachin Sapatnekar. Masoud’s research interests include spintronics circuits and architectures, emerging memory technologies, in-memory computing, computing with post-CMOS devices, 3-D integration, VLSI power distribution network design, and VLSI design automation.
In his doctoral research, Masoud is focused on improving the performance of today’s data processing platforms by developing a spintronics-based true in-memory computing method. The size of the data that must be processed by big data applications is growing exponentially: today’s computational engines are inadequate for the analysis of such large and complex data sets.
With current-day hardware engines struggling to provide solutions to this data onslaught, there is a rapidly growing demand for reducing the gap between the computational requirements of big data applications and today’s computational capabilities. One of the most notable challenges is the large amount of time and energy that is wasted by today’s data processing platforms moving data to and from the memory, where data is stored, and the processor, where computations are performed. Masoud’s spintronics-based method eliminates the access overhead by performing computation inside a memory array. He achieves this through a novel reconfiguration scheme that allows the array to either act as a computational unit, or as a conventional memory unit. Taking this idea from concept to practical implementation requires interdisciplinary work with aspects of materials science/physics, circuit design, and computer architecture. His proposed approaches and platforms are demonstrated to tremendously reduce the energy and execution time required to perform big data computations.
Masoud received his bachelor of science degree from University of Tabriz in 2010, and his master’s degree from Sharif University of Technology in 2013. Both degrees were in electrical engineering and electronics. He is currently pursuing his doctoral degree with the Department of Electrical and Computer Engineering at the University of Minnesota Twin Cities. He has interned with Cadence Design Systems (Voltus R&D team, Austin, TX) over fall 2019 and spring 2020. Preciously, Masoud won the best paper award at the 20th ISQED (March 2019) for his work on in-memory computation using spin-Hall magnetic tunnel junctions.