Decision-Focused Learning Polytechnic Institute Academic Year 2025 Accepted RDE Decision-focused learning (DFL) represents an emerging approach that enhances quality by integrating machine learning (ML) with constrained optimization, training models end-to-end. This paradigm is particularly promising for complex combinatorial decision-making tasks in real-world applications characterized by uncertainty, where accurately estimating unknown parameters for traditional decision models is often a significant hurdle. A prime example of such tasks is vehicle routing, particularly in last-mile and food delivery scenarios. Our current research addresses these challenges using operational data from major companies like Amazon and Meituan. The objective is to develop models whose decision-making capabilities emulate those of experienced drivers/couriers. These human experts implicitly account for numerous dynamic factors, including warehouse logistics, real-time road conditions, and package characteristics (type, size). To achieve this, we are investigating approaches such Decision Transformers, Imitation Learning, and the integration of logical reasoning within deep learning frameworks. Tho Le Support building optimization/ML models and coding; Conduct a literature review; support writing a scientific paper https://thovle.weebly.com/ Any student has min GPA 3.2. Priority is given to students having a strong background in optimization, ML/AI, and logistics/transportation; majoring in industrial engineering/technology, electrical and computer engineering/technology, and computer science/information technology. Students should be strong at communication/speaking, technical writing, and coding (Python) skills. 0 10 (estimated)