We are building the next generation of autonomous AI agents that enable Amazon customers to seamlessly discover and purchase products across the open web. Our systems operate in highly dynamic real-world environments, navigating thousands of third-party merchant experiences with production-grade reliability, scalability, and safety.
This role sits at the intersection of Agentic AI, LLMs, reinforcement learning, multimodal reasoning, and large-scale distributed systems. As an Applied Scientist, you will help define the science and architecture powering internet-scale web agents and autonomous purchasing systems.
In this role, you will:
- Work on autonomous AI agents operating in open-world environments rather than static benchmarks.
- Solve challenging research problems at the intersection of LLMs, web interaction, multimodal reasoning, and scalable systems.
- Influence products that directly impact millions of Amazon customers.
- Build systems that must generalize across thousands of constantly evolving third-party websites.
- Partner closely with science, engineering, and product leaders to shape the future of AI-powered shopping experiences.
- Have the opportunity to publish research, file patents, and contribute to Amazon-wide AI innovation.
Key job responsibilities
- Design scalable evaluation and benchmarking systems for autonomous agents operating in dynamic web environments.
- Develop techniques for robust agent planning, error recovery, and adaptation under distribution shift.
- Build multimodal AI systems that reason over screenshots, DOM structures, user intent, and interaction trajectories.
- Lead scientific direction for agent reliability, task completion, and customer trust.
- Mentor scientists and engineers on advanced AI methodologies and experimentation.
About the team
The Buy For Me purchasing team owns the mission of completing customer purchases on off-amazon sites through both deterministic protocols and agentic methods. This role is with the agentic purchase team which is comprised of both SDEs and Applied Scientists working closely on production agents and offline capabilities to explore off-amazon sites, validate purchasing quality and onboard new sites to the Buy For Me program.
