Visit us at booth #401 in the exhibitor section to say hello and learn more about Nextmv. We’re hosting a workshop, tech tutorial, and session. Learn more below.
Book dedicated time to meet with us at the event.
When: Saturday, October 14, 2023 from 1:00 PM - 2:45 PM
Where: CC-North 122 C
Accessibility: Free and open to all conference registrants
Presenter: Sebastián Quintero
Staff: Nicole Misek, Ryan O’Neil
Isaac Asimov once said, “The most exciting phrase to hear in science, the one that heralds new discoveries, is not ‘Eureka!’ but ‘That’s funny…’” Asmiov’s words speak to the opportunities, curiosities, and showstoppers that surface while building and deploying decision algorithms — but ideally not before an algorithm impacts production systems. That’s where model testing comes in. Testing is a useful tool for algorithm development teams to avoid expensive mistakes, troubleshoot bugs, establish a repeatable path to production, and give stakeholders confidence in results. How does a new model perform compared to an existing one? How do solutions change as constraints are added or removed? Does the model perform as I’d expect before deploying to production?
In this workshop, we’ll explore these questions in addition to walking through the different types of model testing, a framework for bringing them all together, and hands-on examples for how to use and incorporate an optimization model testing toolkit into your OR operations.
View on the INFORMS Annual 2023 website.
When: Sunday, October 15 from 10:45 AM - 11:20 AM
Where: CC-North 120 D
Accessibility: Free and open to all conference registrants
Presenter: Sebastián Quintero
Staff: Nicole Misek, Ryan O’Neil
When optimization technology works well, it feels magical. But it is not magic. Good decision oDecision models save money. Decision tools save time. For decades, realizing business value from decision algorithms and operations research has often been hindered by the challenges with model definition, solver setup, testing, deployment, and collaboration. It’s time for decision optimization technology to get out of its own way. Inspired by MLOps and software development approaches, a new platform is providing collaborative, opinionated tooling that empowers teams to move faster with less confusion and more access to the decision technology ecosystem. In this tech tutorial, we’ll explore these tools and workflows and their impact on accelerating algorithm development cycles from months to weeks or less.
View on the INFORMS Annual 2023 website.
When: Tuesday, October 17 from 2:15 PM - 3:30 PM
Where: TBD
Accessibility: Free and open to all conference registrants
Session 1: The Sushi is Ready, How Do I Deliver It? Developing, Deploying and Managing Real-Time Optimization Systems
Author: Sebastián Quintero
We propose a holistic approach towards real-time optimization systems (such as food & groceries delivery, ride sharing) that transcends traditional OR boundaries. We assert that real-time planning is not an isolated analytical problem, but an integral part of a decision "supply chain", deeply interconnected with other decision models.
We treat these systems as engineered software components in a production stack. We emphasize the necessity of rigorous testing and seamless integration with other software systems. This leads to a more resilient, adaptive, and real-time planning process.
Session 2: Adapting to Change in On-Demand Delivery: Unpacking a Suite of Methodologies
Author: Nicole Misek, Ryan O’Neil
The volatile world of on-demand, gig-economy delivery employs a diverse suite of testing methodologies to master change and operational risk. We explore common testing techniques, in the context of a multi-sided marketplace.
We discuss historical testing, batch experimentation, acceptance testing, shadow testing, switchback testing, and digital twin simulation. We present these not as isolated tools, but as a cohesive toolbox deployed against real-world examples.
View on the INFORMS Annual 2023 website.
