Applied Scientist - Agentic AI, Amazon Fulfillment Technology AI Science

Sourced from emploive.com

Lead the development of agentic AI systems for Amazon's fulfillment network, focusing on multi-agent orchestration, tool use, and memory management. Train and optimize LLMs using advanced reinforcement learning techniques like RLHF, DPO, and SFT to enhance reasoning and decision-making capabilities. Conduct rigorous offline and online evaluations to ensure system reliability and performance in operational environments.

• Build full agentic systems leveraging multi-agent orchestration, tool use, memory, and action execution.
• Train LLMs using rejection sampling, SFT, continual post-training, and Reinforcement Learning (RL).
• Develop training and preference data including reasoning trajectories and tool traces.
• Implement reward modeling and policy optimization (DPO, IPO, RLHF/RLAIF with PPO/GRPO).
• Manage agentic memory/state management, episodic/semantic memory, and grounding with RAG.
• Lead research projects, mentor interns, and author academic papers.

• PhD or Master's degree with 4+ years of CS, CE, ML experience, or equivalent industry experience.
• 3+ years of building models for business applications.
• Proficiency in Java, C++, or Python.