(Image/Midjourney)
Google has awarded a $450,000 gift to several faculty members in the Daniel J. Epstein Department of Industrial and Systems Engineering (ISE) at the USC Viterbi School of Engineering, backing projects that range from artificial intelligence to transportation routing and public health. The awards, announced this year, recognize work that is rigorous, practical and aimed at solving complex problems beyond the lab.
The gifts, which come with access to Google’s specialized computing hardware known as Tensor Processing Units, or TPUs, and its open-source JAX software framework, are highly sought after, with university researchers throughout the world submitting proposals.
“This highlights that we operate at the forefront of AI engineering and applications,” said Meisam Razaviyayn, the Andrew and Erna Viterbi Early Career Chair and associate professor of industrial and systems engineering, computer science, and electrical and computer engineering, who helped coordinate the joint effort. “It signals a strong track record of collaboration with industry.”
Silicon Valley’s tech titans have repeatedly supported ISE. Along with Google, Meta and Amazon have made major gifts to department members.
The five awards were the result of collaboration. Faculty members worked together, presenting Google with a broad vision that stretched across artificial intelligence and several sectors, including transportation, health care, mathematics and education.
The funding will largely support doctoral students so they can focus full time on research. It also provides access to advanced computing systems that can handle problems too large for most university labs.
Here is a closer look at the five funded projects:
Making AI more reliable
Razaviyayn and Karthyek Murthy, an ISE assistant professor, received funding to make artificial intelligence systems more efficient and more reliable. The USC team is working on methods that would allow powerful models to run on more modest hardware, thereby lowering the barrier to entry.
Training today’s most advanced AI models can require staggering amounts of computer memory. That puts cutting-edge research out of reach for many universities and smaller companies. Team members are working on developing resource-efficient approaches to to train and deploy these models using a fraction of the standard memory requirements.
Team members are also tackling another problem: an AI system can be very good at certain benchmarks but still offer poor advice. Accuracy on benchmarks alone does not guarantee sound decisions. The researchers aim to develop tools that help correct those shortcomings without forcing developers to rebuild models from scratch, a process that can cost millions of dollars.
If successful, the work could improve AI-driven decisions in areas such as health care, supply chains and public services.
Untangling traffic
Maged Dessouky, Tryon Chair in Industrial and Systems Engineering at ISE, received funding for a project rooted in a daily frustration for millions of L.A. commuters: traffic congestion.
The project aims to improve the real-time routing of large vehicle fleets. Coordinating thousands of vehicles across a city as complex as Los Angeles is a puzzle of enormous scale. With Google’s computing support, the team wants to make those calculations faster, more efficient and practical.
Improved traffic predictions can also significantly enhance routing decisions which depend on data — lots of it. Ride-hailing services, delivery companies and freight operators all collect valuable information about routes and travel times. However, they are reluctant to share raw data with competitors.
Dessouky’s team is building a system that allows companies to contribute to better traffic forecasts without handing over sensitive details. Instead of sharing the data itself, organizations would share carefully filtered insights, protecting their competitive information while still improving citywide models.
Razaviyayn serves as a co-investigator on the project.
Predicting disease before it spreads
Sze-chuan Suen, an ISE associate professor specializing in health policy modeling, received funding to strengthen the tools used to forecast infectious diseases.
When COVID-19 began spreading, public health officials relied heavily on computer simulations to guide decisions about school closures, hospital capacity and other measures. Many of those tools, however, struggled to keep up with the speed and complexity of a real-world outbreak.
Suen’s project aims to develop user-friendly software that allows researchers, many of whom are physicians and policy experts rather than computer scientists, to run large-scale simulations more quickly. Faster models mean officials can test more scenarios and respond more confidently to emerging threats.
Her team will also provide training to help other researchers adopt the new tools, extending the project’s reach well beyond USC.
Solving hard math problems faster
Giacomo Nannicini, an ISE associate professor, landed funding to work on a tough kind of math problem that regular computer methods struggle to solve.
These problems show up in important real-world situations, such as reducing how much energy a data center uses, improving a financial model, or fine-tuning a manufacturing process. The tricky part is that you can adjust a setting and see what happens, but there’s no simple formula that explains how the change caused the result. Engineers call these “black-box” problems because the inner workings aren’t clear.
Nannicini plans to improve a free software tool he created, called RBFOpt, which helps solve these kinds of problems. He will rewrite parts of it to use powerful computing tools from Google (JAX and TPUs), making it much faster and smarter. The upgraded tool should find high-quality solutions more quickly, even when the data is messy or the problem involves many variables.
Overall, this work improves how we solve complex optimization problems, with practical benefits for energy systems, finance, and engineering design.
Teaching the next generation of AI engineers
All the research breakthroughs in the world mean little if the next generation of engineers does not know how to use the tools behind them. Bruce Wilcox, an ISE senior lecturer and director of MS in Analytics education at USC Viterbi, received funding to address that gap.
Wilcox will develop a complete curriculum centered on JAX, including lab exercises, lecture slides and tutorial videos, and integrate it into graduate-level machine learning and analytics courses at USC. Students will compare JAX with systems they already know, such as TensorFlow and PyTorch, and learn where Google’s framework offers real advantages. All materials will be released publicly on GitHub so universities everywhere can benefit.
Published on April 9th, 2026
Last updated on April 9th, 2026