Loading...
Sr. Applied Scientist, Applied AI Solutions
Amazon
On-siteseniorPosted May 5, 2026
About the Role
Partner with cross-functional teams to identify and deliver high-impact agentic use cases, design, develop, and evaluate long-running agents, and define evaluation frameworks.
Requirements
Excited about pushing past single-turn chat into agents that run for hours or days, designing harnesses, memory systems, and context strategies for reliable AI collaborators on real business problems.
Full Job Description
The newest business group in AWS, Applied AI Solutions are built by AWS and AWS Partners to deliver applied AI solutions that leverage Amazon's operational expertise and that businesses love and trust for their day-to-day success. Our ambition is to become a partner which companies can rely on to run their business every day, putting AI to work delivering better customer experience, operational excellence and speed.
Are you excited about pushing past single-turn chat into agents that run for hours or days, planning, recovering from failure, and accumulating knowledge as they work? Do you want to design the harnesses, memory systems, and context strategies that turn frontier models into reliablen collaborators on real business problems? Do you enjoy the prospect of solving agent problems that haven't been solved at scale anywhere before, including memory decay, belief revision, fact grounding, long-horizon credit assignment, and evaluation of open-ended trajectories?
Along the way, you'll get opportunities to be a disruptor, innovator, and a problem solver, someone who truly enables AI agents to create significant impacts.
Key job responsibilities
You will partner with cross-functional business and engineering teams to identify and deliver high-impact agentic use cases across the Applied AI Solutions portfolio. You will design, develop, and evaluate long-running agents — including orchestration harnesses, memory architectures (episodic recall, semantic facts, revisable beliefs, principled decay), context compaction strategies, and safe, auditable tool and environment designs.
You will define evaluation frameworks for agents whose outputs defy single-answer judgment, building trajectory-level evaluations, reward-hacking detection, and human-in-the-loop review patterns. As a senior scientist, you will set technical direction, mentor scientists and engineers, and represent the science org in roadmap and architecture decisions.
You will ensure seamless deployment and integration of agents into production systems customers rely on daily.
The ideal candidate thrives in ambiguity, builds collaborative relationships, moves fast, and is passionate about customer experience. They hold a strong point of view on where agent capabilities are heading, with the scientific rigor to separate what works from what merely demos well. They lead by example, communicate effectively across audiences, and love solving complex problems at the frontier of applied AI.
Are you excited about pushing past single-turn chat into agents that run for hours or days, planning, recovering from failure, and accumulating knowledge as they work? Do you want to design the harnesses, memory systems, and context strategies that turn frontier models into reliablen collaborators on real business problems? Do you enjoy the prospect of solving agent problems that haven't been solved at scale anywhere before, including memory decay, belief revision, fact grounding, long-horizon credit assignment, and evaluation of open-ended trajectories?
Along the way, you'll get opportunities to be a disruptor, innovator, and a problem solver, someone who truly enables AI agents to create significant impacts.
Key job responsibilities
You will partner with cross-functional business and engineering teams to identify and deliver high-impact agentic use cases across the Applied AI Solutions portfolio. You will design, develop, and evaluate long-running agents — including orchestration harnesses, memory architectures (episodic recall, semantic facts, revisable beliefs, principled decay), context compaction strategies, and safe, auditable tool and environment designs.
You will define evaluation frameworks for agents whose outputs defy single-answer judgment, building trajectory-level evaluations, reward-hacking detection, and human-in-the-loop review patterns. As a senior scientist, you will set technical direction, mentor scientists and engineers, and represent the science org in roadmap and architecture decisions.
You will ensure seamless deployment and integration of agents into production systems customers rely on daily.
The ideal candidate thrives in ambiguity, builds collaborative relationships, moves fast, and is passionate about customer experience. They hold a strong point of view on where agent capabilities are heading, with the scientific rigor to separate what works from what merely demos well. They lead by example, communicate effectively across audiences, and love solving complex problems at the frontier of applied AI.