UC San Diego Jacobs School of Engineering Doubles Size of Early-Career Faculty Acceleration Program

Article Content

The University of California San Diego Jacobs School of Engineering has doubled the size of its Accelerating Interdisciplinary Research Collaborations for Early-Career Faculty program, which it launched last year. The big-picture goal of the program is to help early-career faculty build interdisciplinary research collaborations to the point that they are competitive for multi-year research funding. 

In particular, the program provides funding that enables graduate students from two different labs to begin new research collaborations. At least one of the two UC San Diego Jacobs School of Engineering faculty must be an early-career professor. By funding research collaborations that link two different disciplines, the program also provides exciting research experiences to graduate students. 

This year, six teams won funding. This translates to twelve faculty and twelve graduate students working together to build new research collaborations that will be competitive for external funding. The project is funded by Irwin and Joan Jacobs. 

“It’s not enough to recruit the best faculty. You also have to ensure that early-career faculty get off to a great start,” said Irwin Jacobs, a founding faculty member of UC San Diego and co-founder of Linkabit and later Qualcomm, where he served as founding chairman and CEO. 

Last year, the UC San Diego community celebrated the 25th anniversary of the naming of the Irwin and Joan Jacobs School of Engineering at UC San Diego, in recognition of Irwin and Joan Jacobs’ transformative vision and philanthropy. 

“I am honored to announce that we have doubled the footprint of this innovative program that we launched just last year. With this program, we have a powerful mechanism for accelerating the research efforts of early-career faculty while also nurturing the collaborative culture of the Jacobs School among both our faculty and our graduate students,” said Albert P. PIsano, Dean of the UC San Diego Jacobs School of Engineering and Special Adviser to the Chancellor. “Irwin and Joan Jacobs, thank you for your vision and generosity. You are truly empowering us to prime the pump, when it comes to accelerating the research programs of our early-career faculty.”

The program is currently funded to run for an additional two years beyond this year’s cohort. In total, nearly 50 Jacobs School faculty and graduate students will go through the program, which funds graduate students from different labs over the course of two consecutive quarters. 

The funded collaborations span a wide range of areas, often connecting disparate disciplines in unexpected ways that could open up new areas of research capable of solving big challenges facing society, including the safety, trustworthiness, robustness and efficacy of machine learning and AI systems, and improving indoor air quality while reducing energy costs. 

“Despite the fact that we doubled the number of awards this year, it was incredibly difficult to make the final decisions because we had so many strong applications. I really appreciate everyone who submitted proposals. The appetite for forging new cross-disciplinary research collaborations here at the Jacobs School is truly extraordinary,” said Javier E. Garay, Associate Dean for Research and Professor of mechanical and aerospace engineering at the UC San Diego Jacobs School of Engineering.

This year’s six winning proposals  

Agile Safety: Learning-Enabled Reachability for Safe Multi-Agent Control
Sylvia Herbert, mechanical and aerospace engineering professor
Sicun Gao, computer science and engineering professor

Formal analysis and verification methods provide a mathematically rigorous framework for analyzing the behavior of complex systems, such as autonomous vehicles, space robots, and medical devices. The goal is to derive complete safety guarantees or identify all critical design flaws of the systems, before they are ever deployed in the real world. However, these analysis and verification techniques are almost always too computationally intensive for complex systems in practice. The collaborating graduate students supervised by professors Herbert and Gao will confront these challenges by developing novel methods that allow for lightweight, data-driven, yet principled utilization of formal analysis on practical robotic systems.

Explainable Machine Learning: Survey Experiments and Experiment-Informed Theories 
Parinaz Naghizadeh, electrical engineering professor 
Kristen Vaccaro, computer science and engineering professor

Machine learning models make increasingly important decisions related to health care access, employee hiring, money lending and much more. Today’s machine learning systems, however, are often biased. One of the leading solutions to understand the reasons behind this shortcoming in AI systems involves explainability – which aims to maintain trust and establish accountability for machine learning systems by explaining the reasons behind their recommendations. The collaborating graduate students supervised by professors Naghizadeh and Vaccaro are part of a larger research project aimed at addressing shortcomings in AI explainability by developing guidelines for the design of future explainable AI methods that are human interpretable and economically feasible. 

Building AI: A Foundational AI Model for Modeling and Managing Indoor Environments for Healthy and Energy-Efficient Buildings and Classrooms 
Jingbo Shang, computer science and engineering professor 
Yuanyuan Shi, electrical and computer engineering professor 

The COVID-19 pandemic underscored just how important air quality is in indoor environments such as classrooms and commercial buildings. Maximizing the intake of fresh air and drastically reducing air recirculation for 24-hour seven days a week operation ensures high indoor air quality, but increases energy consumption of buildings by two to three times above nominal energy costs. In addition to the financial burden, this strategy increases both carbon emissions and wear on mechanical systems. The collaborating graduate students supervised by professors Shang and Shi are building the technical foundation for smart building operations, including a foundational AI model for building dynamics modeling, and a constrained optimization approach for optimal building control. The goal is to provide quantifiable safety and comfort guarantees while significantly reducing energy consumption.

Principled Data-driven Control for Societal Nonlinear Systems via Koopman Operator and Behavioral Theory
Yang Zheng, electrical and computer engineering professor
Jorge Cortes, mechanical and aerospace engineering professor

Despite high profile successes like ChatGPT (language processing) and AlphaGo (game playing), many modern data-driven and reinforcement learning techniques suffer from a lack of efficiency and trustworthiness. Today’s successful reinforcement learning systems require millions to billions of data points, days to weeks of training time, but do not provide explicit accountability for safety and robustness. The graduate students supervised by professors Zheng and Cortes are collaborating on an exploratory project aimed at advancing efficiency and trustworthiness in reinforcement learning systems. They are investigating principled data-driven control techniques by leveraging Koopman operator and behavioral theory. The team plans to test their techniques in the control of societal nonlinear systems, such as mixed autonomy. 

High-resolution Quantum Correlated Confocal Microscope
Abdoulaye Ndao, electrical and computer engineering professor
Yeshaiahu Fainman, electrical and computer engineering professor

This team aims to unlock microscope miniaturization, building on complementary electrical engineering and computational skills. The engineers are working to develop a synergic quantum spatial sensing and imaging platform with high-resolution quantum-correlated confocal microscopy. The first planned application will be the quantitative characterization of living cells with unprecedented 3D spatial resolution. The graduate students collaborating on this project are advised by electrical engineering professors Ndao and Fainman. The team’s approach is based on the use of a nano-manufactured photonic nanostructure acting as a multifunctional nanodevice. 

End-to-End Differentiable Wave Optics Simulation for Computational Imaging
Nicholas Antipa, electrical and computer engineering professor
Tzu-Mao Li, computer science and engineering professor

Computational imaging systems rely on algorithms to construct images from data. Real-world use of these systems is varied, with applications in medical imaging, smartphone cameras, virtual and augmented reality and autonomous vehicles. The graduate students led by professors Antipa and Li are collaborating as part of a larger effort to address long-standing issues in computational imaging. The team is working to significantly improve the quality of both the physical imaging system and the image reconstruction algorithm, by accounting for the wave nature of light using a differentiable wave optics simulator and modern data-driven optimization. The key challenges are the large computational cost of the differentiable wave optics simulation, and the discontinuities caused by the pupil and the boundary of the lens system. Solving these challenges could lead us to higher-resolution and crisper images even under high degrees of magnification, and could open new application areas for computational imaging, including the creation of miniature microscopes for imaging single neurons of freely behaving animals.