DoorDash is shifting from legacy one-shot predictions to an agentic recommendation platform, leveraging language-native consumer memory, RQ-VAE semantic IDs, and grounded search to dramatically boost relevance and conversion metrics. The talk by Sudeep Das showcases how to build context-aware consumer AI at scale.
SynthePulse Insight · AI deep readingMembers
Version 1 · 1 source
At QCon AI, DoorDash's Head of Machine Learning, Sudeep Das, shared how the company is shifting from one-shot predictions to an agentic recommendation platform, using language-native memory, semantic IDs, and grounded search to boost relevance and conversion.
At QCon AI 2026, Sudeep Das, Head of Machine Learning and AI at DoorDash, discussed how the company is transitioning from traditional one-shot prediction models to an agentic recommendation platform. This shift aims to address the challenges of consumer AI at scale, particularly in personalization, search, and catalog intelligence.
According to InfoQ, Das shared how this transformation leverages language-native consumer memory, RQ-VAE semantic IDs, and grounded search to significantly improve relevance and conversion. Together, these technologies form a context-aware recommendation system that better understands user needs.
This article is based on InfoQ's coverage of a QCon AI talk, which is a secondary source. The content was provided by a DoorDash executive, but specific data was not disclosed, so claims of metric improvements should be considered as source claims rather than independently verified facts.
Primary report
Primary source