Institution profile

DoorDash

Industry researchnorthamerica · us
Official website
Research library16linked papers
Opportunities35open roles
Selected work

Representative Papers

DashCLIP: Leveraging multimodal models for generating semantic embeddings for DoorDash

Mar 18, 2025arXiv.org

Existing multimodal models struggle to capture fine-grained semantic relationships between merchant items and user queries on the DoorDash platform. Method: We propose a joint multimodal embedding learning framework that requires no user behavioral history. It employs an image–text contrastive learning objective to align pretrained unimodal encoders (CLIP, BERT) with a customized multimodal encoder. To reduce reliance on proprietary business signals, we introduce the first large-scale relevance annotation dataset synthesized via large language models (LLMs). Furthermore, we enable end-to-end joint optimization of both unimodal and multimodal encoders. Results: Experiments demonstrate substantial improvements in item classification and relevance prediction. In advertising recommendation, the method achieves 12.3% lift in click-through rate (CTR) and 9.7% lift in conversion rate (CVR), validating its cross-task generalization capability and direct business impact.

2 citationsRead paper
Recent publications

Latest Papers

Power-Optimal Covariate Adjustment for Switchback Experiments

Jul 29, 2026

In switchback experiments with unequal cluster sizes, between-unit variability inflates the variance of treatment effect estimators, thereby limiting statistical power. Conventional CUPAC methods fail to optimally reduce this variance because they do not distinguish between between-unit and within-unit noise. This work proposes a covariate adjustment approach explicitly designed to maximize statistical power by decomposing the variance structure of the treatment effect estimator and separately optimizing the trade-off between these two noise components during prediction modeling and residualization. We establish the first power-optimal CUPAC theoretical framework that jointly and distinctly accounts for different noise sources, moving beyond the traditional focus on overall prediction accuracy alone. Theoretically, the method achieves maximal statistical power; Monte Carlo simulations confirm its superior variance reduction over standard CUPAC and clarify the practical efficiency gains and applicability boundaries.

0 citationsRead paper