Applied Scientist, Core Search

The Amazon Search team's vision is to deliver high quality search results regardless of how customers phrase their search queries. The Core Search team is reimagining search architecture using Large Language Models (LLMs) and building a new LLM stack that powers various Amazon experiences. We are hiring an Applied Scientist to push the science behind this stack, spanning the full model lifecycle from mid-training reasoning models on shopping data to aligning the models with customers on dimensions like helpfulness, trust, and faithfulness.

- Develop personalized multi-modal thinking-LLM techniques that reason about customers, queries, and products.

- Mid-train and post-train large language models on shopping data: domain-adaptive continued pre-training, reinforcement learning shopping reasoning traces, and instruction tuning for natural-language shopping queries.

- Align models with customer interests on the dimensions such as helpfulness, harmlessness, and faithfulness. Apply Reinforcement Learning (RLVR, RLHF), Direct Preference Optimization (DPO), and customer-behavior-derived reward models.

- Create semantic representations of products, customers, and context (bi-encoder embeddings, contrastive learning, hard-negative mining, cross-lingual training).

- Develop cross-attentive LLM rankers that score candidate products against rich query intent and complex constraints.

- Train multi-objective ranking and optimization systems that balance relevance, purchasability, and personalization.

- Drive improvements on offline benchmarks as well as online experiments.