Research
Decisions about how to use our resources happen every day, from logistics and transportation to energy. I study how AI can augment, not replace, classical operations research to make those decisions better.
Advances in AI and machine learning open new ground for data-driven forecasting, scheduling, optimization, and sequential decision-making. My research integrates AI/ML with operations research to design decision-making pipelines that are both efficient and sustainable, so that practitioners can deploy and generalize intelligent decision systems at scale. I am particularly interested in applied AI for urban mobility and logistics, finding sustainability opportunities while maintaining service efficiency.
Current focus
Generative AI for combinatorial optimization
Fusing generative AI with operations research to rethink how we solve hard combinatorial problems. A recent example is ViTSP (ICLR 2026), which uses vision-language models to guide the search for large-scale traveling salesman problems. A broader theme is treating optimization as a compute-allocation problem: deciding where a finite solver budget will do the most good.
On-demand operations of shared mobility
Learning-based models for planning and operating shared mobility systems, such as deep reinforcement learning for large-scale, online bike-share rebalancing.
Life-cycle carbon footprint of AI systems
Measuring and reducing the environmental cost of AI, including the energy used by neural optimization solvers, so that intelligent systems deliver a net sustainability benefit.