At DoorDash, our expansion into various services like groceries, convenience, and retail goes hand in hand with our commitment to advanced machine learning, especially the integration of Large Language Models (LLMs). In our presentation, we’ll highlight how LLMs are revolutionizing our approach to personalization, search optimization, and product knowledge graph building. These models are instrumental in managing complex scenarios, such as substituting low-in-stock items, and are pivotal in refining our product catalog and taxonomy.

We’ll provide an in-depth look at how LLMs, alongside traditional machine learning techniques, are enhancing user experience. This includes facilitating rapid basket building, simplifying the discovery of new products, and ensuring seamless order fulfillment. Our discussion will focus on the significant impact of LLMs, detailing the lessons learned, challenges encountered, and the practical implications of applying these state-of-the-art technologies in real-world applications at DoorDash.