Hüsrev Cılasun, Alireza Khataei, and Siliang Zeng awarded doctoral dissertation fellowships

Doctoral candidates Hüsrev Cılasun, Alireza Khataei, and Siliang Zeng are recipients of the Graduate School’s 2024-2025 doctoral dissertation fellowship (DDF) award. The fellowship gives the University's most accomplished doctoral candidates an opportunity to devote full-time effort to an outstanding research project by providing time to finalize and write a dissertation during the fellowship year.

Hüsrev Cılasun is conducting his doctoral research under the guidance of Professor Ulya Karpuzcu, and is exploring novel and unconventional ways of computing with spin, the angular momentum of physical particles. He is working on two distinct aspects of such computation. The first one deals with using actual magnetic spins to enable several orders of magnitude higher performance and energy efficiency than conventional computers, while preserving data privacy and fault tolerance for emerging big data problems in, for instance, genomics. The second aspect uses abstract models of spins to solve large scale combinatorial optimization problems which are present in numerous applications in our lives, ranging from robotics, airline scheduling, logistics (determining ideal routes for package delivery) to chip design (finding the best way to draw a chip layout). All of these problems are hard to solve and characterized by several constraints. Cılasun’s research goal is to explore the design space of spin-based computing systems to solve existing problems more efficiently using significantly less resources (including energy) and to solve new problems that no conventional system can due to physical resource limitations. 

 Alireza Khataei has been working on his research under the guidance of Professor Kia Bazargan. Khataei’s research is located at the intersection of innovative data encoding and highly optimized hardware to perform computations with limited hardware resources. The goal is to accelerate compute-intensive applications such as neural networks and image processing. His work stems from the recent explosion of machine learning applications and the growth in AI computation complexity. These developments are the result of advances in hardware speed and innovations in neural network models. However, both the training of the network as well processing the many millions of user prompts entail massive amounts of computation, which requires the support of hardware that can stand up to the challenge. Hardware accelerators are specially designed computation units inside processors that target specific types of computation and are particularly critical for handling the anticipated growth and complexity of neural networks. Khataei’s research targets hardware acceleration, developing innovative solutions to improve the costs associated with computation in terms of time and power. 

 Siliang Zeng’s research is focused on aligning artificial intelligence (AI) systems with human preferences, context, social norms, and other values. Working under the guidance of Professor Mingyi Hong, Zeng’s work develops a comprehensive framework, including formulation, algorithms, and customization to and for specific applications. This will allow AI systems to effectively learn from humans for a wide range of tasks. He addresses critical challenges such as the effective integration of diverse human generated data to build an aligned system, enabling AI systems to continuously learn and adapt to changing contexts and norms, and transparency of AI systems so users can understand and trust them.