How Machine Learning Powers On-Demand Logistics At DoorDash

Recent years have seen a surge in marketplace companies, where the common algorithmic challenge is efficient matching between two sides of the marketplace. For DoorDash, which focuses on food delivery, this problem is even more difficult because of the three-sided marketplace, where we need to identify the optimal Dasher to fulfill a delivery from a restaurant and bring it to a Consumer. At its core, these are more complex versions of vehicle routing problem(VRP). VRP, in its simplest form is NP hard and the real-time, quick turnaround nature of DoorDash introduces additional challenges: delivery requests come in continuously, Dashers constantly are in movement, and the effects of variance in restaurant operations and real world events (traffic, weather, etc.) have pronounced effects on the solutions. Thus, finding global optimality in real-time becomes furthermore intractable. We leverage various machine learning techniques to intelligently model the decision space and achieve near optimal solutions in seconds. Ultimately, across DoorDash’s tens of millions of deliveries, these techniques have led to shorter delivery times for consumers, higher pay for Dashers, increased income for merchant partners, and a better experience for all sides of the marketplace. This talk highlights ML techniques used by DoorDash to enhance efficiency and quality in its marketplace, and provides a framework for how ML can augment core Operations Research problems like the Vehicle Routing problem. We will share an overview of these techniques and dive deep into: 1. Predicting time points in the life of a delivery: Executing a delivery involves decision making on various time points. When will the order arrive to the consumer? When will the food be ready? When will the Dasher arrive at the restaurant? In this section, we will explain how the different predictions are made and more importantly, how they fit in with the larger logistics engine. 2. ML Techniques for batching multiple deliveries together: Algorithms used in the logistics engine to decide when and how deliveries can be batched together
Who is this presentation for?

Prerequisite knowledge:

What you'll learn?

Gary Ren is a machine learning engineer at DoorDash working on its core logistics engine, where he focuses on core AI problems: vehicle routing, Dasher assignments, delivery time predictions, demand forecasting, and pricing. Previously, Gary worked at Microsoft Bing Core Relevance on search ranking and natural language processing.