Explore a thought-provoking conference talk on integrating machine learning and discrete optimization to solve complex real-world problems. Delve into the advantages of combining these fields, including the potential to create more flexible combinatorial solvers capable of learning tailored solution strategies. Discover how combinatorial optimization can be directly integrated into deep learning pipelines, facilitating decision-focused learning where the training loss is a function of downstream optimization decisions. Learn from Bistra Dilkina of the University of Southern California as she presents her insights at the Deep Learning and Combinatorial Optimization 2021 conference, hosted by the Institute for Pure & Applied Mathematics at UCLA.