Sr. Applied Scientist, Amazon Cyber Threat Intelligence

Amazon Amazon · Big Tech · Annapolis Junction, MD · Applied Science

Senior Applied Scientist role focused on inventing and deploying AI/ML systems for cyber threat intelligence at Amazon scale. Responsibilities include identifying and solving complex threat intelligence problems, extending ML techniques for cybersecurity, and implementing production AI/ML systems for threat detection, analysis, and defense. The role involves building predictive models, graph neural networks, and LLM-powered systems, with a strong emphasis on deploying models into production and influencing across teams.

What you'd actually do

  1. Identify, frame, and solve scientifically-complex threat intelligence problems where no textbook solutions exist—including threat scoring, malware classification, infrastructure clustering, and intelligence automation
  2. Drive the scientific agenda for AI/ML within ACTI by proposing research initiatives, defining success metrics, and securing management buy-in
  3. Extend and invent machine learning techniques for cybersecurity applications, including anomaly detection on noisy data, few-shot learning for emerging threat families, and graph-based reasoning over attacker infrastructure
  4. Design, build, and deploy production AI/ML systems that process threat data at scale—from model training on petabyte-scale security logs to real-time inference serving millions of predictions daily
  5. Partner with ACTI engineering teams to integrate AI/ML models into existing intelligence platforms

Skills

Required

  • PhD in engineering, technology, computer science, machine learning, robotics, operations research, statistics, mathematics or equivalent quantitative field
  • 5+ years of relevant, broad research experience after PhD (or equivalent body of work demonstrating scientific innovation)
  • Experience deploying AI/ML models into production systems with direct, verified customer impact
  • Experience in one or more: NLP/LLMs, graph neural networks, anomaly detection, deep learning, or probabilistic modeling
  • Software development

Nice to have

  • cybersecurity applications
  • threat scoring
  • malware classification
  • infrastructure clustering
  • intelligence automation
  • anomaly detection on noisy data
  • few-shot learning for emerging threat families
  • graph-based reasoning over attacker infrastructure
  • model training on petabyte-scale security logs
  • real-time inference serving millions of predictions daily
  • data pipelines
  • feature engineering
  • model training
  • evaluation frameworks
  • production monitoring

What the JD emphasized

  • publish research at peer-reviewed venues
  • deploy AI/ML models into production systems with direct, verified customer impact
  • NLP/LLMs
  • graph neural networks
  • anomaly detection
  • deep learning
  • probabilistic modeling

Other signals

  • invent and deploy novel AI/ML systems
  • automate threat detection
  • accelerate intelligence analysis
  • enable proactive defense capabilities
  • deploy production AI/ML systems that process threat data at scale
  • real-time inference serving millions of predictions daily