Applied Scientist II, Demand Enablement, Product Analytics and Operations

Description

In this role, you will design and build intelligent multi-agent systems that automate root cause analysis for advertising campaign delivery at scale. You will architect agentic orchestration patterns where specialized sub-agents (campaign diagnostics, deal-level troubleshooting, pacing control) are invoked as composable tools by a reasoning layer that determines which subsystems to query based on the nature of the issue. You will develop hierarchical analysis frameworks that move from daily trend detection to intra-day anomaly isolation, enabling the system to pinpoint when and why delivery degraded rather than relying on static time windows. You will build self-learning feedback loops where the system identifies recurring failure signatures (auction dynamics, pacing anomalies, supply contention), updates its diagnostic knowledge as engineering teams deploy fixes, and retires stale patterns automatically.

Key job responsibilities

• Conduct deep data analysis to derive insights for the business, identify gaps, and uncover new opportunities.
• Develop scalable and effective machine learning models and optimization strategies to solve business problems.
• Run regular A/B experiments, gather data, and perform statistical analysis to optimize advertiser experiences.
• Collaborate closely with software engineers to deliver end-to-end solutions into production.
• Enhance the scalability, efficiency, and automation of large-scale data analytics, model training, deployment, and serving.
• Research and implement new machine learning models and techniques to improve advertising performance.

About The Team

The Demand Enablement, Product Analytics and Operations team builds the diagnostic and intelligence layer for Amazon DSP, the demand-side platform powering Amazon's programmatic advertising business. We own the systems that detect, diagnose, and surface delivery issues across campaigns, giving internal teams and advertisers the visibility to act before problems impact spend. Our product portfolio spans automated troubleshooting platforms, advertiser-facing delivery insights, and AI-powered root cause analysis using multi-agent architectures on foundation models. We are a small, high-ownership team that ships production systems end-to-end, from data pipelines processing billions of bid events to LLM-based agents that reason over complex advertising systems.