Based on your passion and background, you may choose to work in a few different areas: Query understanding, Search relevance and ranking, Generative recommendations, LLM evaluation and AIQA systems, Low-latency and scalable LLM systems, Knowledge graphs, Sequence modeling.

Query understanding: Using cutting-edge AI and LLM-based techniques to understand user intent, refine queries, and support downstream retrieval and ranking.

Search relevance and ranking: Improving search relevance by incorporating signals from user behavior, catalog knowledge, and generative models, including hybrid retrieval and ranking systems.

Generative recommendations: Pushing the boundaries of where generative and traditional models intersect across retrieval and ranking systems; developing scalable feedback and reward modeling approaches for closed-loop learning (RFT).

LLM evaluation and AIQA systems: Building LLM-based evaluation frameworks (e.g., LLM-as-a-Judge, self-critique) to improve the quality and reliability of generative and agentic systems.

Low-latency and scalable LLM systems: Researching techniques to deploy LLMs in high-traffic, latency-sensitive production environments, balancing quality, cost, and latency through cascading, distillation, and selective generation.

Knowledge graphs: Working on graph data management and knowledge discovery over one of the world’s largest grocery catalogs, and integrating structured knowledge with LLM-based reasoning and natural language interfaces.

Sequence modeling: Building temporal models for user behavior prediction.