Applied Scientist II, Partner Science

Amazon Ads is building a >$100BN business, and our 3000+ advertising partners — agencies and tech providers — are strategic growth engines for that ambition. The Partner Science team drives the Advertising Partner flywheel by infusing science-based interventions at every stage of the partner journey: demand generation, partner selection, partner engagement and growth, and partner value, and partner experience measurement. We are looking for an Applied Scientist to join our team and develop ML/AI models and causal inference studies that directly improve how advertisers find, work with, and succeed through partners. In this role, you will design, build, and productionize ML/AI and econometric solutions. You will work on ambiguous, real-world and high-impact problems where neither the problem nor the solution is well-defined, and you will be trusted to operate with growing autonomy while collaborating closely with senior and principal product managers, engineers, data engineers, BIEs, and sales/marketing stakeholders. Key job responsibilities • Design, prototype, validate, and productionize ML models across science domains: Predictive/Supervised (e.g., propensity models, deep learning, reinforcement learning), Causal Measurement (A/B tests, causal inference studies), and Text Analytics/LLMs (signal extraction, model explainability, Gen-AI application). • Independently own one or more production science models end-to-end— from initial scoping and design, to final deployment and ongoing monitor and refinement. Conduct scientific literature reviews, benchmark state-of-the-art approaches, and develop novel techniques when no textbook solution exists for our partner ecosystem challenges • Design and run A/B experiments using our scalable experiment framework to validate whether science interventions and product features drive partner growth and ad spend, working with our small, skewed partner population • Perform hands-on data analysis with large-scale advertising datasets, leveraging our centralized Partner Knowledge Base with hundreds of numeric features and unstructured data and Andes data infrastructure Collaborate with the MLOps engineering team to deploy models, and with Data Engineering to create curated datasets that power science and analytics efforts • Translate model outputs into business impact for cross-functional stakeholders (Different Sales and Marketing teams, Finance, Partner Product teams), simplify and provide business friendly communication to drive effective debates and trade-off discussion, and lead to alignment and decision. • Contribute to model quality monitoring, data quality frameworks, and operational excellence — including defining evaluation thresholds and data quality checks at each pipeline stage • Mentor teammates and contribute to a culture of intellectual integrity, continuous learning, and knowledge sharing About the team The Partner Science team sits within the Partner Analytics organization in PartnerTech, Amazon Ads. Our mission is to drive the Advertising Partner flywheel by infusing science-based interventions at all stages of the partner journey — demand generation, partner selection, partner engagement and growth, and partner value — ultimately improving the partner-managed advertiser experience. We are part of a broader Partner Analytics team comprising Data Engineering, Business Intelligence, and Science functions, all unified by a shared commitment to both advertiser and partner success. The Science team currently includes senior applied scientists, data scientists, supported by MLOps engineering partners who help us scale model deployment. Together, we own 10+ production science models and studies that power Partner Network platform features, sales and marketing programs, and finance attribution and forecasting across 20+ marketplaces. We bias for action, embrace a culture of fast iteration and reinforcement learning, celebrate both achievements and lessons learned, and invest in growing top scientist talent. If you are enthusiastic about applying ML/AL, causal inference, and LLMs to real-world advertising ecosystem problems with measurable business impact, we'd love to hear from you. Too learn more about us, see our wiki https://w.amazon.com/bin/view/AdSales/SPE/PEG/Analytics/Science/Overview Basic qualifications - 3+ years of building models for business application experience - PhD, or Master's degree and 4+ years of CS, CE, ML or related field experience - Experience programming in Java, C++, Python or related language - Experience in any of the following areas: algorithms and data structures, parsing, numerical optimization, data mining, parallel and distributed computing, high-performance computing - 3+ years of hands-on predictive modeling and large data analysis experience Preferred qualifications - Experience in professional software development - Ph.D. in computer science, machine learning, engineering, or related fields, or experience in data science, machine learning or data mining - Knowledge of data engineering pipelines, cloud solutions, ETL management, databases, visualizations and analytical platforms - Experience completing complex tasks quickly with little to no guidance and react with appropriate urgency to situations that require a quick turnaround, or experience in sales/business development - • 5+ years experience building propensity models, recommendation systems, or advertiser/partner targeting models in an advertising or marketplace context - • Depth and hands on design experience on A/B test and causal and economic impact analysis. Hands-on experience with LLMs and Gen-AI applications (e.g., text summarization, retrieval-augmented generation, digital assistants) - • Experience with AWS data and ML infrastructure (e.g., SageMaker, S3, Redshift, Athena, Andes, Batch) or equivalent cloud-based ML platforms - • 3+ years experience working in advertising and retail ecosystem, and has done global level projects to unders