About the job
Amazon Live is building the future of shoppable video — connecting brands with customers through livestreams, short-form video, and creator-driven content across Amazon Shopping, Fire TV, Prime Video, and social platforms. The product serves millions of monthly viewers, processes billions of events daily across 9 marketplaces, and generates tens of millions of dollars in advertiser revenue through self-service and managed channels with a goal to reach 100MM+ MAUs and 20K+ brands in the next couple of years. Amazon Live has a mature data platform powering reporting across 9 marketplaces, 14+ dashboards, and real-time creator analytics. The opportunity ahead is different: proving the causal value of video to brands, Amazon's programmatic systems, and product decision-making. This requires production ML models, data experimentation frameworks, and content intelligence that do not exist today — and that is exactly what this role builds.
You will be one of the first scientists on this team — defining the measurement methodology, experimentation standards, and model architecture from the ground up. The problems are high-ambiguity, the data is rich, and the impact is visible — your models will directly influence how brands invest and how millions of customers discover content. Are you excited by the challenge of building causal measurement, content intelligence, and ranking signals from scratch for a product customers interact with daily? Do you want to own the full lifecycle from research question through production deployment, where your work moves multi-million dollar business decisions?
We are looking for an Applied Scientist to join the DESA team and build production-grade models, experiments, and signals that close these gaps. You will own problems end-to-end — from framing the research question through model deployment and A/B experimentation — working alongside Data Engineers who build the infrastructure and a BIE who owns executive reporting. Your outputs will not sit in notebooks. They will run in production, feed downstream ranking systems, power brand-facing metrics, and give partner teams the evidence they need to prioritize integrations with Amazon Live.
Responsibilities
- Design and deploy causal attribution models (incrementality testing, multi-touch) replacing heuristic approaches, producing defensible numbers for partner teams and brand-facing ROI metrics.
- Build brand lifecycle models (LTV, cost-to-acquire, adoption funnel) and campaign optimization models (marketing mix, diminishing returns) that scale self-service revenue.
- Design and run A/B experiments with proper methodology (holdouts, pre-registration, power analysis) for new product surfaces, ranking changes, and attribution model transitions.
- Build multimodal and generative models for content intelligence — extracting structured signals from video and producing scored creative assets at scale.
- Develop ranking and personalization features (content affinity, creator quality indices, cross-session engagement patterns) consumed by downstream distribution systems.
- Build predictive models proving video value to Amazon's programmatic systems where existing retail signals fail.
- Own the full lifecycle from research question through production deployment, monitoring, and iteration.
- Present findings and methodology to senior leadership (Director/VP) and partner teams, translating model outputs into business decisions.
- Contribute to the science community through internal publications, reading groups, and cross-team methodology sharing.
Qualifications
Minimum
- 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 working with PyTorch or JAX software, or experience with Machine Learning and Large Language Model fundamentals, including architecture, training/inference lifecycles, and optimization of model execution
- Experience with A/B testing, especially around audience segmentation and targeting
- Experience with large scale distributed systems such as Hadoop, Spark etc.
- Experience with causal inference methods (incrementality testing, difference- in-differences, instrumental variables, or synthetic control).
Preferred
- Experience in search advertising, search marketing, performance advertising, or similar digital advertising
- Experience with video and image processing and compression algorithms and standards, computer vision and/or machine learning
- Experience in a marketing focused role including customer lifecycle marketing, segmentation reporting, customer funnel analysis, and top-line sales performance
- Experience working with cross-functional teams across business development, marketing, operations, product development, legal, etc.
- Track record of deploying models that directly influenced product decisions or business strategy.
- Experience with Amazon internal tools (Bedrock, SageMaker, Redshift, Cradle) is a plus but not required.