NSF Renews USC-Led AI Optimization Institute With $20M

The National Science Foundation has committed $20 million to extend the Artificial Intelligence Institute for Advances in Optimization (AI4OPT) through 2031, according to details first reported by USC Viterbi School of Engineering. The renewal significantly expands the University of Southern California's role in the multi-institution collaboration with Georgia Tech and UC Berkeley.

USC will receive $5.6 million over five years—more than double its first-phase allocation—and will grow its research team from two to five faculty members. Computer science professor Bistra Dilkina will serve as co-principal investigator alongside Georgia Tech's Pascal Van Hentenryck and Berkeley's Alper Atamtürk.

Why it matters

As critical infrastructure becomes more complex and volatile, traditional optimization methods struggle to keep pace with real-time demands. AI4OPT's work addresses a practical bottleneck: teaching machines to solve optimization problems in milliseconds rather than hours, enabling power grids to adapt to renewable energy fluctuations and supply chains to respond to disruptions. The institute's track record—250 publications and 12 technology transfers in its first phase—demonstrates a path from academic research to deployed systems.

Four research thrusts

The renewed institute organizes its work into four integrated areas. Two focus on advancing core science, while two apply those advances to real-world systems.

Dilkina will lead the AI for Optimization thrust, developing "learning to optimize" approaches and foundational models for discrete optimization. The goal is to create decision-making systems that respond to changing conditions in milliseconds, from power grid management to complex logistics.

The Optimization for AI thrust, led by Berkeley's Atamtürk, will develop techniques to make AI training more scalable and energy-efficient, potentially enabling more robust models that can run on resource-constrained devices.

Two application-focused thrusts will tackle energy systems and supply chains. The energy work addresses the growing complexity of electrical grids as renewable energy and data center demand increase. The supply chain research aims to build more resilient systems for logistics, production, and semiconductor manufacturing that can withstand disruptions from disasters, pandemics, and other global shocks.

Expanded USC team

USC's five-member team spans the Viterbi School of Engineering, Stevens School of Computing and AI, and Marshall School of Business. Computer science professor Willie Neiswanger will focus on uncertainty modeling within AI for Optimization. Industrial and systems engineering professors Andres Gomez and Karmel Shehadeh will lead work on Optimization for AI and agile logistics, respectively. Data sciences professor Vishal Gupta will address methodologies at the intersection of AI and optimization with applications to supply chains.

The university will also host the Seth Bonder summer camp in computational and data science for high school students, extending the institute's educational reach to Los Angeles.

The first phase of AI4OPT produced more than 250 publications, resulted in 12 technology transfers, and created new datasets including Distributional MIPLIB. At USC, the institute helped launch the first PhD specialization and certificate in AI plus Optimization, supported by a separate NSF Research Traineeship grant.

Details of the renewal were announced by USC Viterbi School of Engineering.