Filipe Cabral

Operations Researcher
American Airlines

Bio

Filipe Cabral is an Operations Research professional at American Airlines, working on optimization solutions for the airline industry. He holds a PhD in Operations Research and has previously worked on optimization models for sustainable energy systems. His presentation at INFORMS Analytics 2025 will explore a GPU-accelerated approach to the Crew Pairing Problem, demonstrating how cutting-edge computing can revolutionize traditional airline optimization. Filipe has a strong academic background with publications and presentations at conferences such as INFORMS, ISMP, and ILAS.

Rethinking Crew Pairing Optimization: A GPU-Accelerated Approach

Airlines heavily rely on the Crew Pairing Problem (CPP) to minimize crew costs by creating flight sequences (pairings) that cover all scheduled flights while adhering to complex operational and regulatory constraints. Formulated as a set partitioning problem, CPP’s vast solution space challenges traditional methods like branch-and-price and column generation to their computational limits. These algorithms can even become computationally infeasible for large legacy airlines.
Our research approaches this challenge by leveraging the power of GPUs, the very hardware behind the recent success of Large Language Models and Deep Neural Networks. Specifically, we shift from the solution of computationally expensive linear systems inherent in Simplex and Interior Point methods to a more scalable approach using a first-order method called cuPDLP. Preliminary results using real-world airline data demonstrate significant performance gains over traditional CPU-based approaches, highlighting the potential of our method.

Professional

Filipe Cabral

Operations ResearcherAmerican Airlines

Rethinking Crew Pairing Optimization: A GPU-Accelerated Approach

Get the Latest Updates

Operations Researcher
American Airlines