AI Benchmarking Spec. - Chinese, International Seller Growth

The Seller AI team within International Seller Services organization is focused on helping sellers with the right set of Gen-AI/LLM powered tools and agentic solutions that can enable them to accelerate business growth on Amazon. Our primary focus lies in handling annotations for training, measuring, and improving Artificial Intelligence (AI) and Large Language Models (LLMs), enabling Amazon to deliver a superior seller experience to our sellers worldwide. The AI Benchmarking Associate supports the evaluation of AI systems by designing and executing benchmarking and audit activities to assess model quality, compliance, robustness, and fairness. The role combines elements of AI auditing, quality assurance, and traditional audit-style documentation and stakeholder communication. By joining us, you will play a pivotal role in shaping the future of selling on Amazon for sellers worldwide.

Key job responsibilities

As part of your role, you will have the opportunity to,
• Assist in planning and executing benchmarking exercises for AI models, including defining test plans, metrics, and acceptance criteria across accuracy, robustness, bias, and reliability
• Support content accuracy, relevancy, and privacy checks by reviewing datasets, model outputs, and data handling practices, escalating potential regulatory risks.
• Validate data based on specific annotation guidelines, ensuring the accuracy and quality of the collected information
• Prepare clear audit and benchmarking reports, including error ratings, root-cause analysis, and recommendations, and contribute to presentations for senior stakeholders
Maintain organized audit documentation, evidence, and benchmarking datasets to support internal review
• You will work closely with your team members and managers to drive process efficiencies and explore opportunities for automation
• You will strive to enhance the productivity and effectiveness of the data generation by contributing to the development and continuous improvement of AI audit methodologies, checklists, and test frameworks as regulations and best practices evolve