12-06, 18:30–19:00 (UTC), General Track
What models do you need to run an on-demand logistics operation? Whether you’re building apps for delivery, mobility, or ecommerce, these three decision models can get you started: forecasting, scheduling, and routing. In this talk, we’ll build, test, and deploy each model using Python and Google OR-Tools in a DecisionOps workflow. This talk is for data scientists and decision algorithm developers.
“We need to plan the best shifts to meet demand each week and optimize delivery routes every 30 seconds. Can you do it?” More and more data science teams are tasked with this on-demand business problem. These teams often reach for Google’s OR-Tools, a popular open source software suite for solving optimization problems. However, it’s less common for these teams to be fully versed in the details of decision science and operations research techniques needed to solve it.
In this session, we’ll look at how to build, test, and deploy 3 decision models — forecasting, shift scheduling, and vehicle routing — critical to operating an on-demand logistics business using Python and OR-Tools. Together, we’ll walk through working code for each decision model, review relevant testing techniques to de-risk production rollout, cover data partitioning and model management, and deployment workflows. By the end we’ll have our own working logistics backend.
This talk is for data scientists and decision algorithm developers who are both new to or familiar with Google OR-Tools and decision optimization.
No previous knowledge expected
Ryan is CTO and Co-founder of Nextmv (nextmv.io). Before that, he was an optimization wizard and led the Decision Engineering departments at Zoomer and Grubhub. He studied Operations Research at George Mason University and has a cat.