Customer-obsessed science
Research areas
-
November 11, 20265 min readStatic analysis of deployed networks helps reduce the risk that sensitive data leaks across partitions, supporting the security and reliability of global-scale networking.
-
-
-
-
October 20, 20263 min read
Featured news
-
2026Large Language Model (LLM) based conversational systems generate vast interaction logs containing critical diagnostic signals such as failure modes and safety violations. However, the cross-turn dependencies, implicit state, and topic drift inherent to these conversations (Yang et al., 2025b; Mo et al., 2024) defeat dense retrieval, whose lone embedding cannot concurrently encode discrete structured constraints
-
NeurIPS 2026 Workshop on Large Language Models for Social Reasoning and Simulation (SocialAgent)2026The bitter lesson holds that general methods leveraging computation defeat hand-encoded domain knowledge. We argue it carries a condition its popular reading omits: in every case Sutton cites, the general method was scaled against the same task the domain engineering was solving. Social simulation substituted across tasks instead, replacing hand-built agents with an artifact optimized for predicting text
-
NeurIPS 2026 Workshop on Transitioning from Pre-Training to Post-Training2026Reward fusion combines an exact outcome signal with a dense but noisy process signal before group relative policy optimization (GRPO). For a fixed rollout group, we show that the part of the local GRPO update due to reward is an exact sum of pairwise policy score differences weighted by normalized reward gaps. This identity separates two roles for process reward: it can order incorrect trajectories, but
-
Unified Search and Recommendation Workshop (USRW'26)2026Recommending a new content type to users who have never tried it is a hard cold-start problem: there is no in-domain history to infer taste from. We study audiobook discovery on a large music streaming platform, where most users have never played an audiobook. Many of them, however, leave behavioral traces in other domains: podcasts, audiobooks elsewhere, videos, and related content. We turn those traces
-
PHUSE US Connect 20262026Producing Tables, Listings, and Figures (TLFs) remains one of the most important and expert-dependent steps in clinical study reporting. In a 10-week proof of concept, AI engineers from AWS and statistical programmers from Bayer jointly build a zero-shot, multi-agent workflow that generates structured work packages, executable R code, and draft TLFs with automated checks and human review. Developed on a
Collaborations
View allWhether you're a faculty member or student, there are number of ways you can engage with Amazon.
View all