Build on-demand logistics apps with Python, OR-Tools, and DecisionOps
summary
Three concrete models get built, tested, and deployed: LAD regression for forecasting (robust L1 fit implemented as a linear program), mixed‑integer scheduling that minimizes over/under‑staffing against targets like 3/4/5 orders per driver‑hour, and OR‑Tools routing using the CP‑SAT/local‑search routing library for an 8‑vehicle capacity VRP. Code runs in Python, produces JSON inputs/outputs and cloud visuals, solves instances in seconds, and the DecisionOps workflow pushes code to a remote runtime so outputs and schedules actually materialize.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:53Audience needs
- 3:20Python and OR-Tools for optimization
- 5:25Three core models
- 7:32Forecasting model structure
- 13:14Forecasting demo
- 17:04Scheduling model design
- 20:35Scheduling demo
- 22:22Routing fundamentals
- 25:13Routing demo
- 25:56Summary
related talks
20mRoute Optimization Engine for the Real World
39mSolving the vehicle routing problem for optimizing shipment delivery
32mTips and tricks: data science prototype into production
29mDecision sciences in Python
23mOptimising organized transport at scale
31mMLOps for High-Stakes Environments
32mMLOps at Renault: A Generic Pipeline for Scalable Deployment
15mComputational Mobility: Learning, Prediction, and Algorithms