The North America Stores GenAI Evaluation Media (GEM) team is seeking an Applied Scientist to help shape the future of visual shopping experiences. We're building new CXs and foundational capabilities to understand, enhance, and generate high-quality GenAI imagery, videos and CXs that inspire customers and drive purchase confidence, towards our vision to be the leader in visual media and tools. You will develop machine learning, computer vision and generative AI solutions that create beautiful, accurate, and impactful imagery and videos at scale.
Required Qualifications:
PhD or MS in Computer Science, Computer Vision, Machine Learning or related field
Strong background in deep learning, particularly in computer vision and generative AI
Preferred Qualifications:
PhD in Computer Science, Machine Learning or related fields
Experience with multimodal understanding and foundation models
Background in image/video synthesis, 3D computer vision, or visual quality assessment
Track record of scientific publications in top-tier conferences (CVPR, ICCV, NeurIPS)
Familiarity with e-commerce applications of computer vision
You will help build foundational science solutions that enhance and transform how millions of customers visually evaluate products across Amazon's global stores. Your work will directly influence the shopping experience by creating more engaging, accurate, and personalized visual content that helps customers make confident purchase decisions
Key job responsibilities
Design and implement science primitives for media understanding, standardization, and generation that power multiple visual media types across Amazon's stores
Develop novel approaches for automated quality assurance for maintaining high visual quality standards
Partner with foundational model building teams to leverage and enhance foundation models for visual synthesis and evaluation
Build evaluation frameworks and metrics to measure visual quality, realism, and customer impact
Collaborate with cross-functional teams to integrate science solutions into production systems
