Cyrus Neary Wins 2025 Oden Institute Outstanding Dissertation Award

The Oden Institute for Computational Engineering and Sciences has named Cyrus Neary the recipient of its 2025 Outstanding Dissertation Award. Neary's dissertation, "Engineering AI Systems and AI for Engineering: Compositionality and Physics in Learning," addresses one of today’s most urgent engineering questions: how to transform advances in artificial intelligence into safe, efficient, and trustworthy autonomous systems.

"I am tremendously honored to have received this award," Neary said. "I'm also tremendously grateful to the [Computer Science, Engineering, and Mathematics] CSEM program, the research community at the Oden Institute, my dissertation committee, my advisor Professor Ufuk Topcu, and my colleagues and collaborators—without whom much of my research would not have been possible."

At the core of Neary's research is the challenge of engineering artificial intelligence (AI)-driven systems that not only perform well, but also adhere to the safety, data, and operational constraints typical of real-world applications. His work, which combines deep learning with principles from control theory, physics, and engineering, aims to design algorithms that are simultaneously data-efficient, interpretable, and robust.

"My dissertation work aims to address these questions primarily in the context of data-driven modeling and autonomous operation of robotic systems," Neary said. His research spans two central themes: compositionality—breaking complex problems into manageable subproblems—and the integration of prior physical knowledge into deep learning algorithms.

The first half of Neary’s dissertation introduces new methods for designing compositional AI systems, particularly in the context of sequential decision-making. These methods allow independently developed learning modules to be combined and deployed together, leading to reductions in computational requirements and improved real-world performance. Experiments with robotic platforms, including a real-world unmanned ground vehicle, demonstrated the practical effectiveness of these techniques.

In the second half, Neary focuses on embedding physical structure into machine learning models, particularly for control. By leveraging known mathematical properties of physical systems, he designed control-oriented neural networks that remain stable and effective even when training data is extremely limited. One key result: a learned controller for a hexacopter drone that performs well under real-time conditions, despite minimal data.

According to Neary's advisor, Professor Ufuk Topcu, "Every chapter of Cyrus’s dissertation has been published at top-tier peer-reviewed venues. The work he started in his chapters has continued to mature, and additional publications are still in the pipeline. Cyrus’s dissertation not only solves some interesting problems but also establishes new directions for research that are already beginning to attract attention and will likely stay relevant for the foreseeable future. I believe this feature is what separates outstanding dissertations from good dissertations."

Topcu also praised the practical impact of Neary’s work, noting that "Cyrus pushed innovative algorithmic work into real-world demonstrations on aerial robots in my lab and ground robots with the Army Research Lab. It is rare for researchers to have attained Cyrus’s maturity and demonstrated his level of proficiency in all elements—theory, algorithms, and practice—of research at graduation."

He further remarked, "Cyrus has a natural talent to set the stage and tone in his written communication—in fact, all forms of communication—in an engaging yet rigorous manner. The overall dissertation document and individual chapters perfectly exemplify his talent and care."

Throughout his time at the Oden Institute, Neary was deeply influenced by his collaborations and mentors. He highlighted a multi-year partnership with fellow Ph.D. student Franck Djeumou, also of the Center for Autonomy at the Oden Institute, during which they successfully deployed autonomous flight algorithms. He also credits Mustafa Karabag, another student at the Institute, along with a broader network of colleagues, seminar presenters, and faculty mentors for shaping his perspective and sharpening his research focus.

"Some of the most satisfying moments from my Ph.D. came from watching algorithms that had started as abstract concepts years earlier successfully controlling real-world robotic systems," Neary reflected. "The CSEM Ph.D. program emphasizes contributions across theory, algorithm development, and application, and the research culture at the Oden Institute reflects that philosophy."

Neary's work represents a leap forward in bridging theory and practice in AI-powered systems. As he looks ahead to the next chapter in his research career, Neary remains committed to designing AI systems that are not only powerful, but also safe, explainable, and aligned with human needs.

Neary graduated with his Ph.D. in Computational Science, Engineering, and Mathematics in August 2024 and will return to Austin this May for the commencement ceremony. He is currently a postdoctoral researcher at Mila - The Quebec AI Institute, and he will join the University of British Columbia as an assistant professor in the fall of 2025, where he looks forward to continuing to advance his work at the intersection of machine learning, autonomy, and robotics.

The Oden Institute’s Outstanding Dissertation Award recognizes exemplary research that advances computational science and engineering in a significant way. Eligible dissertations must either be submitted between April 1 of the previous year and April 25 of the current year, or have earned the associated degree during the same period. The award honors dissertations that demonstrate originality, technical depth, and potential for long-term impact—standards that Neary’s dissertation met with distinction.