Pallab Haldar is an accomplished SAP HANA Data Architect, Data Modeler, and SAP Generative AI Researcher with over 19 years of experience in enterprise data architecture, advanced analytics, artificial intelligence, and machine learning. He currently serves as SAP Technical Data Architect and SAP GenAI Specialist at Cognizant Technology Solutions, leading large-scale SAP HANA and AI-driven transformation initiatives for global organizations including 3M, Yokohama Tire, Turck Inc., and Delta. His expertise spans SAP HANA, SAP BTP, SAP Datasphere, generative AI, predictive analytics, cloud-native architectures, and retrieval-augmented generation (RAG) solutions. Pallab has successfully architected enterprise-scale data platforms, implemented AI-powered business automation, and driven performance optimization across complex SAP landscapes. He is an IEEE member, holds multiple SAP professional certifications, and actively contributes to the SAP community through technical publications, blogs, and thought leadership on SAP HANA, AI/ML, and enterprise data innovation.

Title: Intelligent Airline Crew Scheduling Optimization: An Implementation-Oriented AI and Operations Research Framework - Abstract: Crew scheduling at airlines needs to manage several aspects at the same time, i.e. crew members’ qualifications, their duty and rest times as well as costs and fatigue. Moreover, irregular operations need to be recovered from. This article presents an implementation-oriented framework combining demand forecasting by means of an LSTM network and XGBoost with assignment by means of Mixed-Integer Linear Programming (MILP) and a Genetic Algorithm as well as fatigue screening by means of a Random Forest and recovery from irregular operations by means of a Reinforcement Learning agent. A synthetic scenario is used for auditable calculation, i.e. a case consisting of 180 operating days, 48 flight legs per day and a crew pool of 70 members. A worked example covering six periods is presented. Ensemble forecasting of the number of required crew members achieves an Mean Absolute Error (MAE) of 1.03 members and Root Mean Square Error (RMSE) of 1.19 members. The cost modeled on the basis of an optimized assignment of crew members to flights amounts to $8,300 per day compared to $9,800 in the case of a non-optimized assignment. The average time needed for recovery of crew members in case of a disruption is reduced from 74 to 46 minutes by a suitable disruption agent. Note that these numbers are not intended to reflect the performance of a specific airline. Rather, they aim to illustrate the functionality and the validation of the equations in detail. The framework underpins legal rules with support for human interaction as well as for temporal as well as for operational validation.