Raghav Saboo

Raghav Saboo is a machine learning engineer whose work spans search, personalization and language models. His 2026 AI Engineer conference biography identified him as a Staff Machine Learning Engineer and Tech Lead for Search & Personalization within DoorDash’s New Verticals business line, focused on scaling agentic and generative AI integrations in search and recommendation systems. Previously, at Amazon, he developed the first generation of large language models and distilled models for Alexa AI’s new language launches. Before Amazon, he worked as a machine learning consultant, building zero-to-one solutions for clients across multiple industries. He has publications in WSDM, SIGIR and RecSys, and holds a master’s degree from Duke University and combined BEng and MEng degrees from Imperial College London.

In his presentation on marketplace discovery at DoorDash, Saboo describes how semantic understanding helps systems distinguish shopper intent and constraints from engagement signals. He presents a pattern of performing expensive LLM reasoning offline and distilling it into representations and smaller models for efficient serving. The approach includes an offline relevance labeler trained from human seed labels, two-stage contrastive retrieval training, and a relevance tower alongside engagement objectives in ranking. He also explains learned hierarchical semantic IDs for catalog relationships, cold starts and inventory-grounded query reformulation; reusable consumer memory expressed as text, embeddings and graphs; and personalized collections generated from that memory and catalog semantics, with existing systems handling item hydration and ranking. He reports team results of a 2.3% improvement in retrieval relevance NDCG, 4–5% gains in ranking MRR, and early pets-category increases of nearly 1% in order rate and 6% in active users.