My research interest lie at the intersection of discrete optimization, machine learning and network design with application in sustainability and social good. Here are some themes that we are currently pursuing:
– Learning-driven algorithm design: how can we design combinatorial algorithms that have learnable components capable of automatically learning effective strategies on distributions of problems. Examples are learning for branch-and-bound for MIP, learning greedy strategies for graph optimization, learning search strategies in planning, among others
– Decision-focused learning: how can we integrate downstream combinatorial decision making tasks in the training ML models, such that forecasted parameters used in downstream optimization lead to solutions with good ground-truth quality. Examples are clustering as a layer, MIP as a layer, submodular optimization as a layer in gradient-descent-based ML models
– Computational Sustainability: use of AI (ML and optimization) to inform some of the most pressing sustainability challenges. Examples include biodiversity conservation and poaching, disaster resilience & critical infrastructure planning (roads, water, etc.), climate mitigation, opioid/stimulant use, fires, land cover mapping, urban planning
A postdoc will be part of the USC Center for AI in Society (https://www.cais.usc.edu/) and become part of a vibrant and interdisciplinary community of researchers, passionate about using AI for some of the most pressing societal and environmental problems, that value data-driven decision making and engagement of shareholders and domain experts. USC also has a large community of AI researchers (https://ai.usc.edu/key-areas-new/).