ecommerce_ops · ecommerce · workflow
DoorDash uses LLM-generated profiles and content embeddings to improve semantic search and recommendations across verticals
DoorDash faced a persistent bottleneck where personalization depended on embedding quality, which in turn depended on data quality — but sparse metadata flattened catalog richness across all verticals. Behavioral co-visitation approaches tried to bypass this dependency but could not capture identity, context, and intent.
How it works
Common implementation structure
How this type of workflow is generally built, generalized across documented cases — not tied to any one vendor's stack. Click any stage to read what happens there. Specific products that implement these stages appear in “Tools commonly seen” below.
Stage 1 · LLM profile generation
LLMs produce standardized narratives for merchants and items covering ingredients, preparation, attributes, and context.
Tools used
LLMsgemini-embedding-001Qwen 3 Rerank modelMetaflowMiniLMGoogle Gemini embeddingsQwen embedding modelstext-embedding-005text-embedding-3-largeOpenAICohere
Outcome
Deploying LLM-generated profiles with content embeddings improved semantic search and homepage discovery: null search rate fell 3.65%, core search CVR rose 0.66%, and generative personalized carousels drove a 2.4% relative increase in homepage order rate.
Results
Time saved+0.164%
Volume+0.0724%
Grounding & classification
Source type: technical build writeup
45 fields verified against source quotes.
content generationenterprise searchpersonalizationragrecommendation systemproduct catalogfailure mode describedmetric backednamed customerproduction runtime claimedtools describedworkflow describedecommerceaccuracy improvementconversion increaserevenue increasethroughput increasetechnical build writeupecommerce opsdata sync enrichmentrag answering