Principal Applied Scientist, Knowledge Graph & Agentive AI

Amazon; Seattle (WA); On-site; USD 199,000 - 269,000; Full time

Job summary

Amazon Development Center U.S., Inc. seeks an experienced scientist to own the scientific strategy for organizational knowledge representation, retrieval, and graph integration.

You will lead efforts in extraction, entity resolution, and data provenance while shaping proactive AI behavior and evaluation With a background in large language models and multiple ML domains, you will mentor junior scientists, design scalable evaluation, and determine when to deploy frontier vs.

Qualifications

• PhD in CS/ML/Statistics or related field, or Master’s with 8+ years applied science experience.
• 10+ years of experience building and shipping ML/AI systems to production users.
• Deep expertise in large language models and at least two of: information retrieval, knowledge representation and graphs, reinforcement learning, agentic system design, or evaluation methodology for generative systems.
• Demonstrated experience setting technical and scientific direction for a team of scientists, including mentoring senior scientists.
• Hands-on proficiency in Python and the ability to prototype independently in a production codebase.
• Track record of publications, patents, or equivalent evidence of original scientific contribution.

Responsibilities

• Own the scientific strategy for how organizational knowledge is represented, kept current, and retrieved: extraction, entity resolution, deduplication, graph structure, and retrieval that unifies graph, semantic, keyword, and temporal search.
• Advance temporal reasoning. Knowledge changes: facts are revised, decisions are reversed, priorities move. Representing what superseded what and when, and preserving the provenance to distinguish confirmed information from inferred information, is among the hardest open problems in this space.
• Define the science of proactive behavior. When is it right for an AI system to interrupt a human? These are precision‑critical problems where a false positive costs far more than a miss, and where the right threshold varies by team and by individual.
• Lead our measurement science. Build evaluation for completeness and correctness across a multi‑component agentic system, converging on a small number of trustworthy primary metrics rather than a sprawl of component scores. Judge honestly when an offline gain is real and when it is an artifact of a sparse dataset.
• Build the data that doesn't exist. The most valuable phenomena in this domain are also the rarest, which makes naturally occurring examples too scarce to learn from. Design synthetic and simulated data pipelines that generate controlled, realistic scenarios so these capabilities can be developed and tested at all.
• Own the learning loop. Turn human interaction into usable training signal, and set the direction for how the system improves from explicit feedback in the near term and from passive observation over the longer term.
• Make the efficiency calls. Decide where frontier models are required and where a smaller domain‑tuned model is sufficient, and build the cost and capacity measurement that makes it a data‑driven decision rather than an opinion.
• Raise the bar across the team. Mentor scientists, review designs, publish where the work merits it, and represent the science externally to customers and to the research community.

Skills

Python programming

Mentoring senior scientists

Publications or patents

Education

PhD in Computer Science / ML / Statistics or related field

Master's degree with 8+ years applied science experience

Tools

Python

Job description

Amazon Development Center U.S., Inc. seeks an experienced scientist to own the scientific strategy for organizational knowledge representation, retrieval, and graph integration.

You will lead efforts in extraction, entity resolution, and data provenance while shaping proactive AI behavior and evaluation With a background in large language models and multiple ML domains, you will mentor junior scientists, design scalable evaluation, and determine when to deploy frontier vs.