Machine Learning Engineer

The role

You will build end-to-end pipelines to extract signals, train custom models, and serve predictions at production scale. You will work on an automated machine learning ecosystem that recalibrates models based on streaming global telemetry data.

What you will do

• Design, build, and deploy online machine learning models to catch evolving fraud vectors in real time.
• Engineer high-frequency time-series features from over 1 trillion behavioral events.
• Maintain and enhance automated model training and deployment infrastructure.
• Write high-performance code to minimize scoring latency at runtime.
• Work cross-functionally with Core Infrastructure, Product Management, and Data Science teams.

Required qualifications

• 4+ years of professional experience building and deploying large-scale machine learning models into high-traffic production environments.
• Strong proficiency in Java or Scala.
• Strong proficiency in Python.
• Practical experience with Databricks and big data processing frameworks like Apache Spark, Apache Flink, or Hadoop.
• Working with NoSQL data stores like Bigtable.
• Deep understanding of statistical modeling, probability, and standard machine learning algorithms (e.g., XGBoost, Random Forests, Neural Networks, and Clustering techniques).
• Ability to reason through data consistency, pipeline failures, and performance constraints in a distributed, multi-tenant cloud environment (GCP).

Nice to have

• Experience explicitly in the fraud detection, risk mitigation, or cyber-security domains.
• Deep knowledge of streaming architectures (e.g., Apache Kafka).
• Familiarity with containerization and orchestration tools like Docker and Kubernetes.
• Familiarity with leveraging AI coding assistants (e.g., Claude Code) to accelerate development and model prototyping

Before you apply

• Final stage candidates may be asked to travel for in-person final round interviews.

About the company

Sift is the AI-powered fraud platform securing digital trust for leading global businesses. Their data network scores 1 trillion events per year to empower more than 700 customers.