This article explores how pairing large language models with deterministic mathematical optimization solvers transforms wealth management. By enabling natural language interaction with sophisticated portfolio models, the approach empowers advisors and clients to simulate, understand, and adjust financial plans effortlessly—bridging the gap between complex mathematical financial models and user-friendly, real-time decision-making.
Part I of this series explains the business motivation behind a large airline company's spill and recapture (SR) model, which is used to optimize flight itineraries and anticipate unmet demand when aircraft capacities change. The article outlines how the model works and why its combinatorial complexity—driven by numerous flight itineraries and operational constraints—necessitates scalable infrastructure. This sets the stage for the cloud-native solution discussed in Part II.
Learn how OSI has helped one of the major US airlines incorporate delay propagation into their turn building process using CPLEX and JuMP, improving customer satisfaction without adding any additional costs.
Read about how OSI built an optimization-powered app to automate assembly line scheduling for a top lift equipment maker, boosting efficiency and integrating seamlessly with their planning system.
A blog on the advanced optimization algorithm that OSI designed and developed for helping North America's largest School bus fleet management system that spans across hundreds of districts in US and Canada
Discover how decision optimization helps school bus operators cut costs, boost reliability, and plan smarter—beyond routing. Learn how OSI drives efficiency across scheduling, maintenance, and fleet strategy.
We compare Python and Rust in a custom parallel Branch & Bound algorithm, analyzing execution time, scalability, and efficiency to show how language choice affects real-world optimization performance.
How leveraging Machine Learning in Branching Variable Selection (BVS) in a Branch and Bound Algorithm improved MIP solve times, and insights into our model implementation.
