Agentic AI Scientist for Fulfillment Tech & LLMs

Amazon Science; Bellevue (WA); On-site; USD 143,000 - 193,000; Full time

Job summary

Amazon Science seeks an Applied Scientist to lead the development of agentic systems that support operational decision making and orchestration from Bellevue, WA. You will build full agentic systems leveraging multi-agent orchestration, tool use, memory, and action execution.

You will train LLMs using rejection sampling, SFT, continual post-training, and RL, creating training data and optimizing policies for LLMs.

Qualifications

• 3+ years of building models for business application experience
• PhD, or Master's degree and 4+ years of CS, CE, ML or related field experience
• Experience programming in Java, C++, Python or related language
• Experience in algorithms and data structures, parsing, numerical optimization, data mining, parallel and distributed computing, high-performance computing

Responsibilities

• Develop agentic systems to assist with operational decision making and orchestration.
• Research multi-agent orchestration, tool use, memory, and action execution.
• Train LLMs using rejection sampling approaches, SFT, continual post-training, and RL.
• Generate training and preference data for specific use cases (reasoning trajectories, tool traces).
• Reward modeling and policy optimization for LLMs: DPO, IPO, RLHF/RLAIF with PPO/GRPO, rejection sampling.
• Supervised fine-tuning on step-by-step trajectories and tool-use traces.
• Verbal Reinforcement Learning and Continual Learning.
• RL for LLMs, Offline RL and off-policy evaluation.
• Agentic memory/state management; episodic and semantic memory; vector search; grounding with RAG.
• Evaluation: developing decision quality metrics, scaling LLM-based evaluations.

Skills

Model building

Java

C++

Python

Algorithms

Education

PhD in CS/ML

Master's degree + 4+ years CS/ML

Job description

Amazon Science seeks an Applied Scientist to lead the development of agentic systems that support operational decision making and orchestration from Bellevue, WA. You will build full agentic systems leveraging multi-agent orchestration, tool use, memory, and action execution.

You will train LLMs using rejection sampling, SFT, continual post-training, and RL, creating training data and optimizing policies for LLMs.