Assistant Professor Yuxin Chen Receives Prestigious NSF CAREER Award
The National Science Foundation (NSF)’s Faculty Early Career Development Program (CAREER) Award is a five-year prestigious award granted in support of early-career faculty who have the potential to serve as academic role models in research and education and to lead advances in the mission of their department or organization. Our own Assistant Professor Yuxin Chen, who joined the University of Chicago’s Department of Computer Science in the Fall of 2019, was awarded a $569,125 CAREER grant to fund his research on adapting AI-driven learning for real-world experimental design.
Chen’s proposed research asks a fundamental question: how can AI systems learn to choose the most informative experiments in settings where every trial is costly, time is limited, and uncertainty is high? Rather than relying on fixed, hand-crafted acquisition rules, his project develops a new framework for active representation learning, where the system learns both how to represent experimental data and how to decide what data to collect next. This approach is designed to make adaptive experimental design more scalable, more flexible, and more effective in real-world settings.
At the heart of this project is a learning framework that treats experiment selection as a trainable problem. Rather than representing experiments in isolation, the system models the evolving state of both the data and the learning model itself, allowing it to reason about what has been learned so far and what information is still missing. As new observations are collected, the model updates this internal state and adapts its strategy for selecting future experiments, creating a feedback loop between learning and data acquisition. By combining representation learning, probabilistic modeling, and sequential decision making, Chen aims to move beyond static heuristics toward data-driven strategies that adapt as conditions change.
“A key idea in this project is that the system doesn’t just represent individual data points—it represents the state of the model together with the data it has already seen,” Chen said. “This allows it to reason about what information is missing and select experiments that are most useful for improving its future performance.”
The broader goal is to accelerate scientific discovery and improve decision making in domains where experimentation is expensive or constrained. Chen’s work will be evaluated across applications such as simulation-based scientific inference, cyber-physical system tuning, and data-driven materials discovery and protein design. These use cases reflect the real-world promise of the project: helping researchers and engineers to efficiently and confidently identify the most useful measurements.
In addition to the technical contributions of his work, Chen plans to develop new curriculum modules that introduce students to AI for real-world experimentation, create tutorial materials for researchers and practitioners, and involve students in interdisciplinary research opportunities through summer labs and hackathons. In partnership with the Data Science Institute and other collaborators, he aims to broaden participation in machine learning by engaging students from underrepresented groups and building pathways into research.
“We want students to think of learning systems as interactive processes, not just static models,” Chen said. “That means understanding how models evolve with data, and how to actively shape that process through better data collection.”
Together, these efforts reflect the spirit of the NSF CAREER Award: supporting early-career faculty whose work advances both research and education while laying the groundwork for long-term leadership in their field.
“This award supports our effort to build learning systems that don’t just analyze data, but help decide what data to collect in the first place,” Chen said. “I’m excited to see how these ideas can translate into real impact across science and engineering.”
To learn more about Chen’s work, visit his website here.
