Research Scientist , Trustworthy Shopping Experience (TSE)

Are you passionate about solving complex business problems at scale through Generative AI Do you want to help build intelligent systems that reason act and learn from minimal supervision If so we have an exciting opportunity for you on Amazons Trustworthy Shopping Experience (TSE) team.

At TSE our vision is to guarantee customers a worry-free shopping experience by earning their trust that the products they buy are safe authentic and compliant with regulations and policy. We do this in close partnership with our selling partners empowering them with best-in-class tools and expertise to offer a high-quality compliant selection that customers trust.

As a Research Scientist I you will bring subject matter expertise with fundamental improvements in at least one relevant discipline (e.g. NLP computer vision representation learning agentic architecture) to contribute to next-generation agentic AI solutions that automate complex manual investigation processes at Amazon scale. You will invent refine and experiment with solutions spanning agentic reasoning self-supervised representation learning few-shot adaptation multimodal understanding and model compression. With guidance from senior scientists you will stay current on research trends and benchmark your results against the state of the art.

You will help design and execute experiments to identify optimal solutions initiating the development and implementation of small components with team guidance. You will write secure stable testable and well-documented production code at the level of an SDE I rigorously evaluating models and quantifying performance. You will handle data in accordance with Amazon policies troubleshoot issues to root cause and ensure your work does not put the company at risk.

Your scope of influence will typically be at the self-level with the possibility of mentoring interns. You will participate in team design and prioritization discussions learn the business context behind TSEs products and escalate problems with proposed solutions. You will publish internal technical reports and may contribute to peer-reviewed publications and external review activities when aligned with business needs.

This role offers a unique opportunity to contribute to end-to-end AI developmentfrom research through productionwith your contributions serving hundreds of millions of customers within months not years.

Key job responsibilities
Contribute to the design and development of agentic AI systems with multi-step reasoning autonomous task execution and multimodal intelligence including feedback and memory mechanisms leveraging reinforcement learning techniques for agent decision-making and policy optimization with input and guidance from senior scientists
Develop novel models built on top of SFT (Supervised Fine-tuning) and RFT (Reinforced Fine-tuning) approaches as well as few-shot approaches based on multimodal datasets spanning text images and structured data applying mathematical optimization techniques to improve efficiency resource allocation and decision-making in complex workflows working alongside senior scientists to identify optimal solutions
Contribute to building production-ready deep learning and conventional ML solutions including multimodal fusion and cross-modal alignment techniques that seamlessly connect visual textual and relational understanding to support automation requirements within your teams scope
Help identify customer and business problems; use reasonable assumptions data and customer requirements to solve well-defined scientific problems involving multimodal inputs such as unstructured text documents product images and relational data developing representations that capture complementary signals across modalities and mapping business goals to scientific metrics
May co-author research papers for peer-reviewed internal and/or external venues including contributions in areas such as multimodal representation learning and vision-language modeling and contribute to the wider scientific community by reviewing research submissions when aligned with business needs
Prototype rapidly iterate based on feedback and deliver small components at SDE I levelincluding multimodal data pipelines and inference modulesthat integrate into production-scale systems
Write secure stable testable maintainable and well-documented code balancing model capability deployment cost and resource usage across multimodal architectures while understanding state-of-the-art data structures algorithms and performance tradeoffs
Rigorously test code and evaluate models across individual and combined modalities quantifying their performance; troubleshoot issues research root causes and thoroughly resolve defects leaving systems more maintainable
Participate in team design scoping and prioritization discussions through clear verbal and written communication; seek to learn the business context science and engineering behind your teams products including how multimodal signals contribute to trust and safety decisions
Participate in engineering best practices with peer reviews; clearly document approaches and communicate design decisions; publish internal technical reports to institutionalize scientific learning
Help train and mentor scientist interns; identify and escalate problems with proposed solutions taking ownership or ensuring clear hand-off to the right owner

- PhD in a quantitative field or MS degree and 1 years of quantitative field research experience
- Experience investigating the feasibility of applying scientific principles and concepts to business problems and products
- Experience using SQL databases to manage and analyze large data sets
- Knowledge of machine learning processing: computer vision NLU NLP or operations research

- PhD in operations research applied mathematics theoretical computer science or equivalent
- Experience applying forecasting and data mining techniques

Our inclusive culture empowers Amazonians to deliver the best results for our customers. If you have a disability and need a workplace accommodation or adjustment during the application and hiring process including support for the interview or onboarding process please visit for more information. If the country/region youre applying in isnt listed please contact your Recruiting Partner.