Researchers from the U.S. National Science Foundation AI Institute for Advances in Optimization (AI4OPT) played a major role in the 2025 INFORMS Annual Meeting, held in Atlanta. With more than a dozen presentations and posters, AI4OPT demonstrated how artificial intelligence and optimization are transforming industries from logistics and energy to e-commerce and transportation.

The meeting brought together operations research and analytics professionals from around the world. Georgia Tech’s Kevin Dalmeijer served as a senior member and arrangements co-chair, contributing to the organization of the large event. The conference concluded with a plenary address from AI4OPT Director Pascal Van Hentenryck, highlighting the institute’s leadership in the global research community.

AI4OPT’s Impact Across Domains

AI4OPT researchers presented innovative work spanning freight logistics, energy systems, semiconductor manufacturing, and renewable forecasting. Many of these studies integrated machine learning and optimization to improve decision-making in complex, data-driven environments.

Tinghan Ye introduced a Contextual Stochastic Optimization framework for omnichannel e-commerce fulfillment, addressing delivery time uncertainty through data-driven modeling. Sikai Cheng presented SPOT, a hybrid optimization and machine learning approach that reduces costs and emissions in freight transportation networks.

In the energy sector, Kevin Wu presented a high-resolution transmission and storage expansion planning framework for renewable heavy power systems, while Alireza Moradi shared a physics-informed conformal prediction method to enhance renewable energy forecasting accuracy. Guancheng Qiu introduced a dual conic proxy architecture for semidefinite relaxation in power flow optimization, providing faster, more scalable computations.

Innovation in Transportation and Logistics

Several presentations focused on real-world logistics challenges. Ved Mohan introduced a framework for resource substitution in large networks, combining machine learning and operations research to ensure equitable resource allocation. Stefan Faulkner’s deep learning framework predicted delivery delays using conformal prediction to enhance reliability and trust in logistics systems.

Yunji Kim presented a practice-based optimization model for locomotive assignment which produced scalable and practical solutions for freight rail operations. Similarly, Jeren Konak developed load consolidation models to improve linehaul efficiency for major parcel carriers.

Other logistics research included uncertainty-aware load planning (Thomas Bruys) and trailer path optimization (Mohammed Faisal Ahmed), each leveraging large industrial datasets to improve tactical decision-making and reduce operational costs.

Other notable AI4OPT contributions included work by Jorge Huertas on semiconductor manufacturing scheduling, highlighting new constraint programming models for resilient production systems.

Bridging Machine Learning and Optimization

Pascal Van Hentenryck delivered the meeting’s closing plenary, “Learning to Optimize: Foundations and Industrial Impact.” The presentation explored how integrating machine learning with optimization can accelerate real-time decision making in complex systems such as energy grids, logistics networks, and supply chains. His talk emphasized how these hybrid approaches transform theory into actionable solutions that drive industrial innovation.

Collectively, AI4OPT’s research showcased the institute’s mission to advance optimization science and deploy scalable AI systems that address pressing industrial and societal challenges.

AI4OPT Presenters at INFORMS 2025

AI4OPT affiliated presenters at the 2025 INFORMS Annual Meeting included:

• Ahmed El-Nashar
• Alan Erera
• Alireza Moradi
• Amira Hijazi
• Andrew Rosemberg
• Guancheng Qiu
• Jeren Konak
• Jorge Huertas
• Kevin Wu
• Mathieu Tanneau
• Mohammed Faisal Ahmed
• Pascal Van Hentenryck
• Reza Zandehshahvar
• Sikai Cheng
• Stefan Faulkner
• Thomas Bruys
• Tinghan Ye
• Ved Mohan
• Yunji Kim

About AI4OPT

The U. S. National Science Foundation AI Institute for Advances in Optimization (AI4OPT) is a National Science Foundation AI Institute led by Georgia Tech. The institute brings together researchers in artificial intelligence, operations research, and engineering to revolutionize how large-scale decision making is optimized for efficiency, sustainability, and resilience.