An Optimization Engine Inside a Freight Platform

The platform’s planning product had outgrown basic planning logic. Real freight operations meant optimizing dispatches for cost or transit time under constraints like pickup and delivery time windows, material compatibility, and multi-stop routing, a class of problems that requires genuine Operations Research depth.

The engine had to serve two very different dispatch models without separate systems: the classic depot-based model where every order ships from one location, and distributed networks where each order can have its own pickup and drop-off anywhere.

Building that expertise in-house would have meant hiring and standing up a specialized OR team from scratch, a significant investment that would have slowed the product roadmap.

We worked as the platform’s Operations Research team, designing and building an optimization-based planning engine alongside its engineers, fitted to its product and its customers’ operations.

The engine solves both core dispatch models used across the industry: single-pickup multi-drop, where every order originates from the same depot, and multi-pickup multi-drop, where a single vehicle can pick up from one customer and deliver to another anywhere on the network.

Dispatch plans respect the constraints real operations run on: pickup and delivery time windows, distance limits, vehicle-route compatibility, and hop-distance limits, among others. Solutions were engineered to run within operational time limits and to be explainable enough for product and customer success teams to reason about.