Dec 3, 2025 · At RecSys 2025, we shared how DoorDash built a large language model (LLM) powered framework that turns restaurant orders and search into cross-vertical affinity features, …

DoorDash addressed the challenge of behavioral silos in their multi-vertical marketplace, where customers have deep interaction history in some categories (like restaurants) but sparse data …

Dec 30, 2025 · Combining LLMs with knowledge graphs and flexible retrieval solves the precision-recall tradeoff in complex search scenarios.

We conducted rigorous set of offline and online experiments to mea-sure the impact of LLM-generated features on our item-ranking model trained on three months of data.

Aug 17, 2025 · I'm excited by DoorDash's recent engineering blog post, “Profile Generation with LLMs: Understanding consumers, merchants, and items” (July 29, 2025), which highlights a …

Mar 8, 2025 · Key insights from Doordash's implementation can be found in the original source. 3 Changes to Expand DoorDash’s Product Search Beyond Delivery. Last updated: March 08, …

Nov 6, 2025 · Our novel approach trains a query encoder with an LLM-curated relevance dataset, eliminating the reliance on engagement history. These embeddings demonstrate strong …

DoorDash's Summer 2025 intern program showcased several sophisticated LLMOps implementations that demonstrate how large-scale food delivery platforms leverage language …

Dec 3, 2025 · A recap of our KDD 2025 PARIS Workshop talk: “Affordability, Familiarity, and Novelty: An LLM-assisted Personalization Framework for Multi-Vertical Retail Discovery.”

Dec 30, 2025 · DoorDash built three LLM-powered pipelines to extract different attribute types. Brand identification matters for sponsored ads and product affinity. DoorDash maintains a …