How DoorDash uses LLM-generated profiles and embeddings to improve semantic search, recommendations, cold-start relevance, and discovery.

DoorDash offers insights into food delivery technology, logistics, and customer experience. Developers can delve into DoorDash's engineering blogs, tech talks, and innovation stories to gain a deeper understanding of how technology powers the food delivery industry. By exploring topics such as scaling infrastructure, optimizing delivery routes, and enhancing user experience through innovative technology solutions, developers can use insights to apply to their own projects and industries.

Doordash

DoorDash engineers describe how they replaced sparse behavioral signals with LLM-generated merchant and item profiles to power content-first embeddings across food, grocery, retail, and gifting verticals. The post covers the full pipeline: incremental Metaflow-based embedding refresh, evaluation using an LLM-as-a-judge harness (Hit@K, nDCG@K), and model selection (gemini-embedding-001 with 256-dim MRL). Key finding: data quality (LLM profiles) dominates over model choice, delivering +31% item similarity gains vs. +6% from a better encoder alone. Production results include a 3.65% reduction in null search rate, +0.66% search CVR, +2.4% homepage order rate, and 68→85% offline precision@10. The post also discusses limitations of text-based consumer embeddings and future directions including semantic IDs, generative retrieval, and context-conditioned consumer representations.

Using LLMs to build content embeddings for DoorDash search and recommendations

Traditional playbook for content embeddings

Limitation: Consumer embeddings from consumer profiles