Senior Applied Scientist, Amazon Ads, Demand Tech , Amazon Advertising, Demand Tech

Amazon Amazon · Big Tech · Palo Alto, CA · Applied Science

Senior Applied Scientist role focused on building and improving deep learning models for response prediction and incrementality in Amazon's advertising platform. The role involves end-to-end ownership from design to production deployment, with a strong emphasis on low-latency, high-throughput inference and online A/B testing. Collaboration with engineers on serving infrastructure and mentoring junior scientists are also key aspects.

What you'd actually do

  1. Own end-to-end response prediction — design and improve deep learning models for multi-task prediction (click, conversion, page view, incrementality) serving at inference latencies under 10ms at millions of TPS
  2. Build and iterate on calibration mechanisms that keep prediction accuracy stable across rapidly shifting supply distributions
  3. Integrate novel signals (OpenRTB features, customer behavioral sequences, supply quality feeds) into production models to improve optimization quality
  4. Run online A/B experiments at scale, analyze results with statistical rigor, and translate offline gains into measurable business impact
  5. Collaborate closely with engineers on model serving infrastructure (SageMaker, GPU inference, real-time feature stores) to deploy models efficiently at scale

Skills

Required

  • building machine learning models for business application experience
  • PhD, or Master's degree and 6+ years of applied research experience
  • Experience programming in Java, C++, Python or related language
  • Experience with neural deep learning methods and machine learning

Nice to have

  • Experience with modeling tools such as R, scikit-learn, Spark MLLib, MxNet, Tensorflow, numpy, scipy etc.
  • Experience with large scale distributed systems such as Hadoop, Spark etc.

What the JD emphasized

  • deep learning models
  • multi-task prediction
  • inference latencies under 10ms
  • millions of TPS
  • online A/B experiments at scale
  • model serving infrastructure
  • GPU inference
  • real-time feature stores

Other signals

  • ML models at massive scale
  • optimize programmatic advertising performance
  • response prediction and incrementality models
  • bid optimization
  • deep learning models for multi-task prediction
  • serving at inference latencies under 10ms at millions of TPS
  • online A/B experiments at scale
  • model serving infrastructure (SageMaker, GPU inference, real-time feature stores)