Four Columbia Engineering Professors Win 2021 NSF CAREER Award

Feb 26 2021

Clockwise: Yuri Faenza, Dan Lacker, Carl Vondrick and Lauren Marbella.

Four Columbia Engineering professors have won the National Science Foundation’s (NSF) Faculty Early Career Development awards this year. The CAREER award is the NSF’s most prestigious award given to support junior faculty with the “potential to serve as academic role models in research and education and to lead advances in the mission of their department or organization.”

Yuri Faenza, assistant professor of industrial engineering and operations research, has received an NSF CAREER award for his proposal to develop an algorithmic theory of matching markets, markets in which agents cannot buy a product just by paying its price. Instead, they receive a product that is a function of both their preferences and the preferences of other agents. His project aims to understand the computational limits of current matching markets, and to devise alternative models and solutions for these markets that currently elude computational tractability, ones for which we know that a good solution exists, but we do not know how to find it efficiently. Faenza is particularly interested in exploring models that have a fundamental societal impact, such as those used to address the lack of diversity in school cohorts and to resettle refugees.

Faenza, an expert in discrete optimization, is focused on developing and using tools from multiple mathematical sources to expand the understanding of matching markets. Matching markets have been a fundamental topic in computer science, economics, and operations research since 1962, when researchers introduced them to model the admission of students to colleges and demonstrated an efficient algorithm that is still used today, with modifications, for assigning students to schools in many cities, including New York.

But as matching markets have evolved in complexity, the algorithms and the related theory have not always been able to keep up. With the NSF award, Faenza will study the underlying mathematical structures in existing models and design new, more efficient algorithms to address the many different contexts of matching markets. These range from enhancing diversity in school cohorts and allocating housing to assigning users to servers and refugees to their new homes.

Through a number of global collaborations, Faenza’s group will test the algorithms on real-world data on refugee resettlement, provided through a non-profit organization that assigns refugees to their new homes. Supporting open-source code, the team aims at uploading their algorithms for anyone who wants to use them on their favorite markets.

Dan Lacker, assistant professor of industrial engineering and operations research, has won an NSF CAREER award for his proposal to develop new mathematical methods for analyzing large-scale competitive systems governed by networks. Lacker works at the intersection of applied probability, stochastic analysis, and mathematical finance and expects his systematic study of network effects to dramatically expand the scope of applications of mean field game theory. This could be especially effective in the mathematical modeling of systemic risk and stability in financial markets as well as the spread of ideas through social networks, pedestrian crowd dynamics, income equality, and much more.

Diverse areas of science rely on mathematical models of large-scale systems of interacting and competing “agents.” Agents may be investors, drivers, or viruses, and the large systems may be financial markets, highways, or even epidemics. The most widely used framework for modeling competition between many interacting agents is "mean field game theory," a method used to model dynamic interactions in which the environment changes in response to agents’ behavior. This technique works well in symmetric models, in which each agent interacts equally with each of the others.

But many systems are governed by networks , such as those formed by social ties or financial obligations that determine which players interact with each other. The complex but important effects of network structure cannot be adequately captured by the standard theory. Lacker’s project tackles this limitation by developing new mathematical frameworks to handle large-scale models of competition with interactions governed by networks.

Lauren Marbella, assistant professor of chemical engineering, has won an NSF CAREER award for her research to improve the stability of lithium metal (Li) batteries and optimize their performance. Li battery technologies are critical to the future of renewable energy resources and electrified transportation and lightweight, low-cost, long-lasting Li batteries would be transformative in mitigating climate change.

There are, however, a number of challenges that have prevented successful commercialization of next-generation lithium batteries. Electrochemical interfaces control the performance of a wide range of applications that include batteries, fuel cells, and sensors. For example, Li metal anodes develop a solid electrolyte interphase (SEI) immediately upon contact with a liquid electrolyte composed of electrolyte decomposition products. Instabilities in the SEI on rechargeable Li metal anodes result in uneven Li deposition architectures and can lead to short-circuits and other safety hazards, including explosions and fires.

The NSF award will support Marbella’s project to use nuclear magnetic resonance (NMR) spectroscopy to explore the specific chemical mechanisms that underpin SEI formation and function, and to then develop methods to achieve the desired smooth Li deposition in these batteries. In situ NMR is a very powerful tool that can directly measure electrolyte decomposition in real time, allowing her group to pinpoint the specific chemical compounds that control performance degradation in electrochemical systems. Marbella, whose research bridges the gap between material design and characterization, is using a combination of solution and solid-state NMR (SSNMR), together with magnetic resonance imaging (MRI), to assess the chemical the chemical structure, arrangement, stability, and dynamics of the SEI on Li metal in situ .

Marbella’s group is highly skilled in materials synthesis, electrochemistry, characterization, and in situ techniques. She is particularly interested in creating functional platforms for a wide range of energy-related materials, including batteries, catalysts, and optics. Her work using SSMNR has already resulted in one of the first molecular-level characterizations of the SEI in Li metal batteries, as well as new insight into how Li ion movement in the SEI is linked to battery performance. This new NSF-funded project will continue her research to advance the way to lightweight, low-cost, long-lasting energy storage for electric vehicles, houses, and much more.

Carl Vondrick, assistant professor of computer science, has won an NSF CAREER award for his research program to develop machine perception systems that robustly detect and track objects even when they disappear from sight, thereby enabling machines to build spatial awareness of their surroundings.

Vondrick’s research focuses on computer vision and machine learning. Over the last four years, he and his team have established a framework for unsupervised machine learning that capitalizes on the large amounts of raw video found on the internet. The computer models were trained without human supervision and, through hours of observing videos, they learned to anticipate human actions, generate short videos, and track objects over time.

This NSF-funded project expands on those results and addresses the fundamental challenges for machines to learn about spatial representations on their own. For example, a child understands that once a toy disappears inside a container, it still exists even though it is not directly visible. Machines, however, are currently not able to detect objects once they disappear from view. The state-of-the-art systems in machine perception are often unable to encode fundamental principles about space that are intuitive to people.

Since spatial awareness underpins complex behaviors in humans and animals, successful machine spatial awareness will also ignite a corresponding generation of capabilities across health, security, and robotics. For example, spatial awareness will enable artificial perception to remain accurate during emergencies where visibility may be poor and objects become obstructed. In robotics, spatial awareness is essential for autonomous vehicles to safely and fluidly interact with people who often become occluded or disappear from view in crowded streets. In other cases, situational awareness and being aware of what is happening in a scene will amplify computer vision for unconstrained scenes, such as busy places of work that are often cluttered and rapidly changing.

Despite rapid progress in deep learning, foundations of visual intelligence are still largely missing from computer vision today. Vondrick aims to address this need and expects his project to generate algorithms and a framework that will fortify computer vision deployments for the real complexity of everyday situations.